04.10.20 · algebraic-geometry / moduli

Deformation theory of smooth curves

shipped3 tiersLean: none

Anchor (Master): Harris-Morrison *Moduli of Curves* (Springer GTM 187, 1998); Sernesi *Deformations of Algebraic Schemes* (Springer Grundlehren 334, 2006); Kodaira-Spencer 1958 *On deformations of complex analytic structures I, II* (Ann. of Math. 67); Kuranishi 1962 *On the locally complete families of complex analytic structures* (Ann. of Math. 75); Schlessinger 1968 *Functors of Artin rings* (Trans. AMS 130); Mumford 1965 *Geometric Invariant Theory*; Deligne-Mumford 1969 *The irreducibility of the space of curves of given genus* (Publ. Math. IHES 36)

Intuition Beginner

A smooth projective curve $C$ of genus $g$ is a one-dimensional object, but the space of nearby curves — curves that look like $C$ after a small perturbation — has many more dimensions. The deformation theory of $C$ is the study of these small perturbations: how many directions can you wiggle the curve, and what controls those directions?

The remarkable answer is that the count of independent wiggles is exactly $3 g - 3$ for $g \geq 2$ , matching Riemann's 1857 count of moduli for the global moduli space $M_{g}$ . The local and global pictures fit together: each point of $M_{g}$ has a tangent space of dimension $3 g - 3$ , and the directions in that tangent space are exactly the first-order deformations of the curve sitting at that point.

The same count specialises to lower genus. A genus-0 curve is the projective line $P^{1}$ and has no moduli at all — every smooth genus-0 curve over the complex numbers is isomorphic to $P^{1}$ , so there is one curve, hence zero parameters and zero deformation directions. A genus-1 curve is an elliptic curve, parametrised by a single complex number called the $j$ -invariant; the deformation count is one. For $g \geq 2$ the count is $3 g - 3$ , and the elementary formula comes from $dim H^{1} (C, T_{C})$ where $T_{C}$ is the tangent line bundle of the curve.

The second key fact is that deformations of smooth curves are unobstructed: every first-order deformation can be extended to higher orders, all the way to a full one-parameter family, with no obstruction blocking the lift. This is what makes the moduli space $M_{g}$ smooth at a curve with no extra automorphisms — and it is a direct consequence of the vanishing of $H^{2} (C, T_{C})$ for a smooth curve, which holds because $T_{C}$ is a line bundle and the second cohomology of a line bundle on a one-dimensional variety vanishes by dimension reasons.

Visual Beginner

A schematic of a smooth projective curve at the centre, with $3 g - 3$ independent arrows pointing outward to nearby deformed curves. Each arrow is one direction of perturbation in the moduli space $M_{g}$ ; together they span the tangent space at the central curve. A second panel shows the matching count from the cohomology side: the tangent line bundle $T_{C}$ on the curve, with $H^{1} (C, T_{C})$ labelled as the moduli tangent space and $H^{2} (C, T_{C}) = 0$ flagged as the obstruction-vanishing that makes the moduli space smooth.

The picture compresses the two central facts of curve deformation theory: the tangent space to $M_{g}$ at a curve is the cohomology group $H^{1} (C, T_{C})$ of dimension $3 g - 3$ , and the obstruction space $H^{2} (C, T_{C})$ vanishes because $C$ is one-dimensional. The first fact identifies the moduli; the second guarantees that the moduli space is smooth at the curve.

Worked example Beginner

Compute the dimensions of $H^{1} (C, T_{C})$ for the three classical genera and verify that they match the moduli counts on $M_{g}$ .

Step 1. Genus 0: $C = P^{1}$ . The tangent line bundle of the projective line is $T_{P^{1}} = O_{P^{1}} (2)$ , a line bundle of degree 2. The first cohomology of $O_{P^{1}} (2)$ vanishes by direct computation: every degree-2 line bundle on $P^{1}$ has no first cohomology. So $H^{1} (P^{1}, T_{P^{1}}) = 0$ , the deformation space is zero-dimensional, and there are no moduli. The moduli space $M_{0}$ is a single point.

Step 2. Genus 1: $C$ an elliptic curve. The tangent line bundle of an elliptic curve is $T_{C} = O_{C}$ , the structure sheaf (because the genus-1 curve has canonical bundle equal to the structure sheaf $O_{C}$ , and the tangent bundle is dual to the cotangent bundle, which equals the canonical sheaf). The first cohomology of $O_{C}$ on a genus-1 curve has dimension $g = 1$ by definition of the geometric genus. So $H^{1} (C, T_{C}) = H^{1} (C, O_{C})$ has dimension 1, the deformation space is one-dimensional, and there is one modulus — the $j$ -invariant. The moduli space $M_{1}$ is the affine line.

Step 3. Genus 2: $C$ a hyperelliptic curve. The canonical bundle of a smooth genus- $g$ curve has degree $2 g - 2$ ; for $g = 2$ this is degree 2. The tangent bundle is dual to the canonical bundle, so $de g T_{C} = - (2 g - 2) = - 2$ for $g = 2$ . By the curve-genus formula, $h^{0} (C, T_{C}) - h^{1} (C, T_{C}) = de g T_{C} + 1 - g = - 2 + 1 - 2 = - 3$ . Since $T_{C}$ has negative degree, the zeroth cohomology vanishes ( $h^{0} = 0$ ), so $h^{1} (C, T_{C}) = 3$ . This matches Riemann's formula $3 g - 3 = 3$ at $g = 2$ , and $M_{2}$ has dimension 3.

Step 4. General $g \geq 2$ . The same computation goes through: $de g T_{C} = 2 - 2 g$ , $h^{0} (C, T_{C}) = 0$ for $g \geq 2$ (the tangent bundle has negative degree, so no global sections), and by Riemann-Roch $h^{1} (C, T_{C}) = - (2 - 2 g) - 1 + g = 3 g - 3$ . So $dim H^{1} (C, T_{C}) = 3 g - 3$ for $g \geq 2$ , matching the Riemann count of moduli.

What this tells us: the cohomological count $dim H^{1} (C, T_{C})$ recovers Riemann's $3 g - 3$ moduli formula on the nose, and the three cases $g = 0, 1, g \geq 2$ are unified by the single formula. The deformation theory of smooth curves provides the local-at-a-point picture that fits together with the global GIT construction of $M_{g}$ .

Check your understanding Beginner

Formal definition Intermediate+

Let $k$ be an algebraically closed field of characteristic zero, let $C$ be a smooth projective curve over $k$ of genus $g$ , and let $T_{C} = Ω_{C}^{1}^{\lor}$ denote the tangent sheaf (the dual of the cotangent sheaf $Ω_{C}^{1}$ , see 04.08.01). Write $Art_{k}$ for the category of Artin local $k$ -algebras with residue field $k$ .

Definition (deformation of a scheme). Let $A \in Art_{k}$ . A deformation of $C$ over $A$ is a flat morphism $C \to Spec A$ together with an isomorphism $C \times_{Spec A} Spec k ≅ C$ identifying the closed fibre with $C$ . Two deformations $(C, ι)$ and $(C^{'}, ι^{'})$ are equivalent iff there exists an $A$ -isomorphism $C \to C^{'}$ commuting with the identifications on the closed fibre.

Definition (deformation functor). The deformation functor of $C$ is $$ \mathrm{Def}_C : \mathrm{Art}_k \to \mathrm{Set}, \qquad A \mapsto {\text{deformations of } C \text{ over } A}/!\sim. $$ A first-order deformation is an element of $Def_{C} (k [ϵ] / ϵ^{2})$ where $k [ϵ] / ϵ^{2}$ is the ring of dual numbers.

Definition (Kodaira-Spencer map). Let $π : X \to B$ be a smooth family of smooth projective curves over a base scheme $B$ , with fibres $X_{b} := π^{- 1} (b)$ . The Kodaira-Spencer map at a closed point $b \in B$ is the connecting homomorphism $$ \rho_b : T_b B \to H^1(X_b, T_{X_b}) $$ arising from the short exact sequence $$ 0 \to T_{\mathcal{X}/B}|{X_b} \to T{\mathcal{X}}|{X_b} \to \pi^* T_B|{X_b} \to 0, $$ where $T_{X / B}$ is the relative tangent sheaf. The map $ρ_{b}$ records how the fibres deform as one moves the base parameter at $b$ .

Definition (versal deformation). A formal deformation $C \to Spf R$ over a complete local Noetherian $k$ -algebra $R$ with residue field $k$ is versal for $C$ iff for every $A \in Art_{k}$ and every deformation $C_{A} \to Spec A$ of $C$ , there exists a $k$ -algebra map $R \to A$ such that the pullback of $C$ along this map is equivalent to $C_{A}$ . The deformation is miniversal (or semiuniversal) iff the differential $m_{R} / m_{R}^{2} \to m_{A} / m_{A}^{2}$ of this map is uniquely determined.

Definition (obstruction theory). Let $0 \to I \to A^{'} \to A \to 0$ be a small extension in $Art_{k}$ (so $m_{A^{'}} \cdot I = 0$ ). The obstruction to lifting a deformation $C_{A} \in Def_{C} (A)$ to a deformation in $Def_{C} (A^{'})$ is a class $$ \mathrm{obs}(\mathcal{C}_A, A') \in H^2(C, T_C) \otimes_k I, $$ whose vanishing is equivalent to the existence of such a lift. The deformation functor $Def_{C}$ is unobstructed iff all obstruction classes vanish.

Counterexamples to common slips

The deformation functor classifies isomorphism classes of deformations, not deformations as pointed schemes. Two deformations of $C$ over $A$ that are abstractly isomorphic as $A$ -schemes inducing the identity on the closed fibre give the same element of $Def_{C} (A)$ .
The Kodaira-Spencer map is not in general an isomorphism — it is an isomorphism exactly when the family $X \to B$ is versal at $b$ . For a non-versal family the Kodaira-Spencer map may be neither injective nor surjective.
Unobstructedness of the deformation functor of $C$ relies on $H^{2} (C, T_{C}) = 0$ , which holds because $T_{C}$ is a line bundle on a one-dimensional variety. For a smooth projective surface or higher-dimensional variety $X$ the obstruction space $H^{2} (X, T_{X})$ can be non-zero, and deformations can be obstructed (Mumford's 1962 example of obstructed deformations of curves in $P^{3}$ exhibits this in the Hilbert-scheme setting).
The moduli space $M_{g}$ is smooth of dimension $3 g - 3$ at $[C]$ exactly when $C$ has finite automorphism group; in particular, $Aut (C)$ must be reduced to the identity for $M_{g}$ to be smooth at $[C]$ as a scheme. For curves with extra automorphisms (hyperelliptic, with extra symmetry, etc.), the Deligne-Mumford stack $M_{g}^{stack}$ remains smooth but the coarse moduli space $M_{g}$ has quotient singularities.

Key theorem with proof Intermediate+

Theorem (Kodaira-Spencer 1958 — classification of first-order deformations). Let $C$ be a smooth projective curve over an algebraically closed field $k$ of characteristic zero. There is a canonical bijection $$ \mathrm{Def}_C(k[\epsilon]/\epsilon^2) \cong H^1(C, T_C). $$ Under this bijection, the zero class corresponds to the unique deformation $C \times_{k} Spec k [ϵ] / ϵ^{2}$ obtained by base change. ^{[Kodaira-Spencer 1958]}

Proof. The argument has three steps. First, build a first-order deformation from an open cover plus Čech cocycle data. Second, identify the cocycle data with $H^{1} (C, T_{C})$ . Third, check that the identification is bijective and compatible with the equivalence relation on deformations.

Step 1: Čech cocycle from a deformation. Let $U = {U_{i}}$ be an affine open cover of $C$ . A first-order deformation $C \to Spec k [ϵ] / ϵ^{2}$ is locally identified with $U_{i} \times_{k} Spec k [ϵ] / ϵ^{2}$ via affine charts, because every affine scheme has only the identity first-order deformation (the deformation functor of an affine scheme is locally identity by infinitesimal lifting of smooth morphisms). The global deformation $C$ is therefore the glueing of the local product deformations $U_{i} \times_{k} Spec k [ϵ] / ϵ^{2}$ along the overlaps $U_{ij} := U_{i} \cap U_{j}$ .

On each overlap, the two local trivialisations differ by an $A$ -automorphism of $U_{ij} \times Spec k [ϵ] / ϵ^{2}$ inducing the identity on the closed fibre. Such an automorphism is given by its action on the structure sheaf: $f \mapsto f + ϵ \cdot θ_{ij} (f)$ for an $O_{U_{ij}}$ -derivation $θ_{ij} : O_{U_{ij}} \to O_{U_{ij}}$ , that is, a section $θ_{ij} \in T_{C} (U_{ij})$ of the tangent sheaf over $U_{ij}$ . The collection $(θ_{ij})$ is a Čech $1$ -cochain of $T_{C}$ on the cover $U$ .

The cocycle condition $θ_{ij} + θ_{j k} = θ_{ik}$ on triple overlaps $U_{ij k}$ comes from compatibility of the glueings around the triangle — the three glueings around a 3-fold overlap must compose to the identity. So the cochain $(θ_{ij})$ is a Čech $1$ -cocycle in $\overset{ˇ}{C}^{1} (U, T_{C})$ .

Step 2: cohomology class is well-defined. Two equivalent deformations differ by a global $A$ -isomorphism $Φ : C \to C^{'}$ restricting to the identity on the closed fibre. Locally on each $U_{i}$ , $Φ$ is given by $f \mapsto f + ϵ \cdot ξ_{i} (f)$ for a derivation $ξ_{i} \in T_{C} (U_{i})$ — that is, a section of the tangent sheaf on $U_{i}$ , forming a Čech $0$ -cochain. The relation between the two cocycles $(θ_{ij})$ and $(θ_{ij}^{'})$ produced by $C$ and $C^{'}$ is exactly the coboundary equation $$ \theta'{ij} - \theta{ij} = \xi_j - \xi_i, $$ so they differ by the Čech coboundary $δ (ξ_{i})$ . The Čech cohomology class of $(θ_{ij})$ in $\overset{ˇ}{H}^{1} (U, T_{C}) = H^{1} (C, T_{C})$ depends only on the equivalence class of the deformation $C$ .

Step 3: bijectivity. The construction gives a map $Def_{C} (k [ϵ] / ϵ^{2}) \to H^{1} (C, T_{C})$ . This map is surjective: given any Čech $1$ -cocycle $(θ_{ij})$ of $T_{C}$ , the formula above produces glueing data for an $A$ -scheme, and the cocycle condition ensures associativity of the glueing on triple overlaps, so the glued $A$ -scheme exists as a first-order deformation of $C$ . Flatness over $Spec k [ϵ] / ϵ^{2}$ is automatic for any glueing of product deformations by tangent-sheaf cocycles. The map is injective by Step 2: equivalent deformations produce cohomologous cocycles, so the equivalence class is determined by the cohomology class.

Combining the three steps, the map $Def_{C} (k [ϵ] / ϵ^{2}) \to H^{1} (C, T_{C})$ is a canonical bijection, and the product deformation $C \times Spec k [ϵ] / ϵ^{2}$ corresponds to the zero cohomology class. $□$

Bridge. The Kodaira-Spencer identification builds toward the construction of the moduli space $M_{g}$ as a scheme of dimension $3 g - 3$ , and the central insight is that the tangent space to $M_{g}$ at a curve $[C]$ with $Aut (C) = {1}$ is the cohomology group $H^{1} (C, T_{C})$ — putting these together, the local-at-a-point deformation theory and the global GIT construction agree exactly on the dimension count. This is exactly the bridge between the analytic Kodaira-Spencer 1958 framework (deforming complex structures on a fixed underlying topological surface) and the scheme-theoretic Mumford 1965 framework (parametrising abstract curves via tri-canonical embedding and GIT quotient). The foundational reason this works is that the Čech-cocycle realisation of a first-order deformation identifies the tangent sheaf cohomology with the glueing data of an infinitesimal family, and the resulting cohomological count is exactly $3 g - 3$ by Riemann-Roch on the tangent line bundle.

This bridge appears again in 04.10.01 (moduli of curves), where the tangent space to $M_{g}$ at a curve with identity automorphism group is identified with $H^{1} (C, T_{C})$ and the GIT-quotient dimension count $dim M_{g} = dim Hilb^{P} - dim PGL$ is matched against $3 g - 3$ , and in the Deligne-Mumford stack $M_{g}^{stack}$ , where the cotangent complex generalises the tangent-sheaf cohomology and the same orbifold tangent space appears. The pattern recurs in deformation theory of higher-dimensional varieties: $H^{1} (X, T_{X})$ is always the tangent space to the deformation functor, and $H^{2} (X, T_{X})$ is always the obstruction space, but for $dim X \geq 2$ the obstruction space can be non-zero and deformations can be obstructed.

Exercises Intermediate+

Exercise 2 (easy, multiple choice).

The Kodaira-Spencer map $ρ_{b} : T_{b} B \to H^{1} (X_{b}, T_{X_{b}})$ for a family $π : X \to B$ of smooth projective curves arises from which short exact sequence?

The Euler sequence on projective space
The conormal exact sequence for a divisor
$0 \to T_{\mathcal{X}/B}|_{X_b} \to T_{\mathcal{X}}|_{X_b} \to \pi^* T_B|_{X_b} \to 0$
The Atiyah class of a holomorphic vector bundle

Hint

The Kodaira-Spencer map records how fibres deform under a tangent-vector move in the base.

Answer

C. The Kodaira-Spencer map is the connecting homomorphism in cohomology coming from the short exact sequence $0 \to T_{X / B} ∣_{X_{b}} \to T_{X} ∣_{X_{b}} \to π^{*} T_{B} ∣_{X_{b}} \to 0$ of locally free sheaves on the fibre $X_{b}$ . The connecting map $H^{0} (X_{b}, π^{*} T_{B} ∣_{X_{b}}) = T_{b} B \to H^{1} (X_{b}, T_{X_{b}})$ is by definition $ρ_{b}$ . The Euler sequence and conormal sequence are unrelated, and the Atiyah class is a different invariant on vector bundles.

Exercise 3 (medium, symbolic).

Use Riemann-Roch on the tangent line bundle $T_{C}$ of a smooth projective curve $C$ of genus $g$ to derive the formula $dim H^{1} (C, T_{C}) = 3 g - 3$ for $g \geq 2$ .

Hint

The tangent bundle has degree $2 - 2 g$ . Apply Riemann-Roch for curves: $h^{0} (L) - h^{1} (L) = de g L + 1 - g$ . For $g \geq 2$ , $de g T_{C} < 0$ so $h^{0} (T_{C}) = 0$ .

Answer

The canonical line bundle $ω_{C}$ on a smooth projective genus- $g$ curve has degree $2 g - 2$ (a classical fact from Riemann-Roch on the canonical bundle, equivalently the degree-genus identity $de g ω_{C} = 2 g - 2$ ). The tangent bundle $T_{C} = ω_{C}^{\lor}$ is the dual line bundle, so $de g T_{C} = - (2 g - 2) = 2 - 2 g$ .

For $g \geq 2$ , $de g T_{C} = 2 - 2 g \leq - 2 < 0$ , and a line bundle of negative degree on a smooth projective curve has no non-zero global sections (otherwise the section would vanish on a non-empty effective divisor of non-negative degree, contradicting negative degree). So $h^{0} (C, T_{C}) = 0$ .

Riemann-Roch on the smooth projective curve $C$ for the line bundle $T_{C}$ reads $h^{0} (C, T_{C}) - h^{1} (C, T_{C}) = de g T_{C} + 1 - g = (2 - 2 g) + 1 - g = 3 - 3 g .$

Substituting $h^{0} (C, T_{C}) = 0$ : $- h^{1} (C, T_{C}) = 3 - 3 g$ , hence $h^{1} (C, T_{C}) = 3 g - 3$ . This recovers Riemann's moduli count for $g \geq 2$ . Rubric: full credit for the Riemann-Roch computation and the substitution.

Exercise 4 (medium, short-answer).

Show that the obstruction to lifting a first-order deformation of a smooth projective curve $C$ to a second-order deformation lies in $H^{2} (C, T_{C})$ , and explain why this group vanishes for any smooth projective curve.

Hint

Repeat the Čech-cocycle construction at second order. The obstruction is the failure of the cocycle condition for a chosen second-order lift, computed as a cohomology class.

Answer

Obstruction class. Let $C_{1}$ be a first-order deformation of $C$ over $A_{1} = k [ϵ] / ϵ^{2}$ , classified by a class $η \in H^{1} (C, T_{C})$ . To lift $C_{1}$ to a second-order deformation over $A_{2} = k [ϵ] / ϵ^{3}$ , choose local lifts: on each affine open $U_{i}$ of an open cover $U$ , lift the local first-order isomorphism to a second-order $A_{2}$ -isomorphism. The chosen local lifts may fail to satisfy the cocycle condition on triple overlaps by a discrepancy in $T_{C} (U_{ij k})$ ; this discrepancy is a Čech 2-cochain $(ζ_{ij k})$ , the obstruction cochain. A direct computation (Sernesi §1.2.5) shows that $(ζ_{ij k})$ is a Čech 2-cocycle, and that changing the local lifts modifies $(ζ_{ij k})$ by a Čech coboundary. Therefore the obstruction is a well-defined cohomology class $obs (C_{1}) \in H^{2} (C, T_{C})$ , and it vanishes iff a second-order lift of $C_{1}$ exists.

Vanishing. For a smooth projective curve $C$ the tangent sheaf $T_{C}$ is a line bundle, and the variety $C$ has dimension one. By Grothendieck's vanishing theorem in dimension one (or directly: the Čech complex of any sheaf on a one-dimensional variety, computed on an affine open cover, vanishes in degree $\geq 2$ ), $H^{i} (C, F) = 0$ for every coherent sheaf $F$ on $C$ and every $i \geq 2$ . Applied to $F = T_{C}$ , this gives $H^{2} (C, T_{C}) = 0$ . Therefore every obstruction class vanishes, every first-order deformation lifts to a second-order deformation, and by iteration every first-order deformation lifts to a formal family over $Spf k [[t_{1}, \dots, t_{3 g - 3}]]$ . Rubric: full credit for the Čech-cocycle obstruction construction and the dimension-one vanishing argument.

Exercise 5 (medium, short-answer).

For a smooth projective genus-2 hyperelliptic curve $C$ given by $y^{2} = f (x)$ with $de g f = 6$ (six distinct roots), describe the moduli space $M_{2}$ as a quotient and verify $dim M_{2} = 3$ .

Hint

A genus-2 curve is hyperelliptic, parametrised by its six branch points on $P^{1}$ modulo $PGL_{2}$ .

Answer

Every smooth genus-2 curve admits a unique 2-to-1 morphism to $P^{1}$ , and this morphism has six branch points on $P^{1}$ (by Riemann-Hurwitz: $2 g - 2 = 2 \cdot (- 2) + R$ gives $R = 2 g + 2 = 6$ ramification points, each contributing one branch point). So the curve is determined by its unordered set of six branch points on $P^{1}$ , modulo automorphisms of $P^{1}$ .

The unordered six-tuples on $P^{1}$ with distinct entries form the open subset $Sym^{6} (P^{1})^{\circ} \subset Sym^{6} (P^{1})$ , of dimension $6$ . The group $PGL_{2} (k)$ of Möbius transformations acts on $Sym^{6} (P^{1})^{\circ}$ , with $dim PGL_{2} = 3$ . The quotient is the moduli space $M_{2} = Sym^{6} (P^{1})^{\circ} / PGL_{2},$ and the dimension count gives $dim M_{2} = 6 - 3 = 3$ , matching $3 g - 3 = 3$ at $g = 2$ .

Explicit local coordinates: normalise three of the six branch points to $0, 1, \infty$ using $PGL_{2}$ . The remaining three branch points are three distinct points $λ_{1}, λ_{2}, λ_{3} \in P^{1} ∖ {0, 1, \infty}$ , providing local coordinates on $M_{2}$ . The deformation-theoretic tangent space at $[C]$ is $H^{1} (C, T_{C})$ of dimension 3, and the three local coordinates $λ_{1}, λ_{2}, λ_{3}$ are exactly the local-parameter realisation of this tangent space. The deformation $de g f = 5$ case is the closure relation: when one branch point moves to infinity, the sextic degenerates to a quintic, and the same moduli are parametrised on a complementary affine chart of $M_{2}$ . Rubric: full credit for the $Sym^{6} (P^{1}) / PGL_{2}$ identification, the dimension count, and the explicit three-coordinate parametrisation.

Exercise 6 (medium, short-answer).

For a smooth quartic plane curve $C \subset P^{2}$ (a non-hyperelliptic genus-3 curve), verify the dimension count $dim M_{3} = 6$ both from the Riemann formula $3 g - 3$ and from the projective-geometry count of plane quartics modulo $PGL_{3}$ .

Hint

A smooth quartic plane curve has 15 coefficients (homogeneous polynomial of degree 4 in 3 variables, up to scalar). The group $PGL_{3}$ acts by linear changes of coordinates. Take the open locus of smooth quartics and quotient.

Answer

Riemann count. A smooth plane quartic has genus $g = (d - 1) (d - 2) /2 = 3$ by Plücker's formula (see 04.05.07). The deformation-theoretic count gives $dim M_{3} = 3 g - 3 = 6$ .

Projective-geometry count. A homogeneous polynomial of degree 4 in 3 variables has $(2 4 + 2) = 15$ coefficients. The projective space of quartic curves on $P^{2}$ is $P^{14}$ (the 15 coefficients modulo scalar), of dimension 14. The open subset $U \subset P^{14}$ of smooth quartics has the same dimension $14$ . The group $PGL_{3}$ acts on $U$ by linear changes of coordinates on $P^{2}$ , with $dim PGL_{3} = 9 - 1 = 8$ . The stabiliser of a generic smooth quartic in $PGL_{3}$ is finite (in fact the identity for a generic quartic, by the canonical-embedding theorem for non-hyperelliptic curves), so the quotient has dimension $dim (U / PGL_{3}) = 14 - 8 = 6,$ matching the Riemann formula. The non-hyperelliptic locus of $M_{3}$ is identified with this open subset; the hyperelliptic locus (where the genus-3 curve has a 2-to-1 map to $P^{1}$ ) is a codimension-one stratum of $M_{3}$ realised separately as a $Sym^{8} (P^{1}) / PGL_{2}$ quotient of dimension $8 - 3 = 5$ , sitting inside the 6-dimensional $M_{3}$ at codimension one. Rubric: full credit for both dimension counts and the matching $dim = 6$ identification.

Exercise 7 (hard, short-answer).

State Schlessinger's criterion for prorepresentability of a deformation functor, and verify that the deformation functor $Def_{C}$ of a smooth projective curve $C$ with $Aut (C) = {1}$ satisfies the four Schlessinger conditions (H1)-(H4).

Hint

Schlessinger 1968 (Trans. AMS 130, 208-222) gives four conditions (H1)-(H4) on a functor $F : Art_{k} \to Set$ that together are necessary and sufficient for $F$ to admit a miniversal pro-representing object. For curve deformations, (H1) and (H2) hold by Čech-cocycle glueing, (H3) is finite-dimensionality of $H^{1} (C, T_{C})$ , and (H4) is equivalent to the rigidity coming from the identity automorphism group.

Answer

Schlessinger's criterion (1968). Let $F : Art_{k} \to Set$ be a functor with $F (k) = {*}$ a one-point set. Consider, for every pair of Artin local $k$ -algebras $A \to A^{''}$ and $A^{'} \to A^{''}$ , the natural map $α : F (A \times_{A^{''}} A^{'}) \to F (A) \times_{F (A^{''})} F (A^{'}) .$ The Schlessinger conditions are: (H1) $α$ is surjective whenever $A^{'} \to A^{''}$ is a small extension. (H2) $α$ is bijective whenever $A^{''} = k$ and $A^{'} = k [ϵ] / ϵ^{2}$ . (H3) The tangent space $t_{F} := F (k [ϵ] / ϵ^{2})$ is finite-dimensional as a $k$ -vector space. (H4) $α$ is bijective whenever $A^{'} \to A^{''}$ is a small extension and $A = A^{'}$ . If (H1)-(H4) hold then $F$ admits a miniversal pro-representing complete local Noetherian $k$ -algebra $R$ , and $F$ is pro-represented by $R$ (in the precise sense that $F = Hom_{k -alg, cont} (R, -)$ ) iff (H4) is upgraded to bijectivity in (H1).

Verification for $Def_{C}$ . Let $F = Def_{C}$ for $C$ a smooth projective curve with $Aut (C) = {1}$ . (H1) and (H2): the Čech-cocycle realisation of first-order deformations gives surjectivity of the pullback map on fibred products of Artin rings, with bijectivity at the tangent-space level — Sernesi §2.3.2. (H3): the tangent space $t_{F} = F (k [ϵ] / ϵ^{2}) = H^{1} (C, T_{C})$ has finite dimension $3 g - 3$ for $g \geq 2$ by the previous theorem; finite for $g = 0, 1$ similarly. (H4): the upgrade from surjectivity to bijectivity for $A = A^{'}$ amounts to showing that an automorphism of an $A^{'}$ -deformation that is the identity on the closed fibre and the identity on the base change to $A^{''}$ is the identity — equivalently, that the automorphism functor of an infinitesimal deformation is rigid. This follows from $H^{0} (C, T_{C}) = 0$ for $g \geq 2$ , which holds because $T_{C}$ has negative degree $2 - 2 g < 0$ and a negative-degree line bundle on a smooth projective curve has no global sections. Therefore the deformation functor $Def_{C}$ is pro-represented by a complete local Noetherian $k$ -algebra of relative dimension $3 g - 3$ over $k$ , and the miniversal family is the formal family over $Spf k [[t_{1}, \dots, t_{3 g - 3}]]$ . Rubric: full credit for the statement of (H1)-(H4) and the verification of each condition for $Def_{C}$ .

Exercise 8 (hard, short-answer).

Describe how the deformation theory of smooth curves extends to the deformation theory of nodal curves, and indicate how the local-versal deformation of a nodal curve enters the Deligne-Mumford compactification $\overline{M_{g}}$ of $M_{g}$ .

Hint

A node singularity $x y = 0$ has a one-parameter local deformation $x y = t$ smoothing the node. The deformation theory of a nodal curve has a node-smoothing direction at each node, plus the standard $3 g - 3$ moduli of the underlying smoothing.

Answer

Nodal-curve deformation. Let $C$ be a nodal curve (a projective curve whose only singularities are nodes — points analytically locally isomorphic to $Spec k [[x, y]] / (x y)$ ). At each node $p_{i}$ of $C$ there is a local versal deformation $x y = t_{i}$ smoothing the node, with $t_{i}$ a local parameter on the deformation base. The global deformation theory of $C$ is a fibred extension of the deformation theory of the normalisation $C \to C$ (a disjoint union of smooth components) by the local node-smoothing parameters. Concretely, the tangent space to the deformation functor of a nodal curve $C$ with $δ$ nodes is $t_{Def_{C}} = Ext^{1} (Ω_{C}^{1}, O_{C}),$ and the local-to-global spectral sequence gives $0 \to H^{1} (C, T_{C}^{ns}) \to t_{Def_{C}} \to i ⨁ T_{p_{i}}^{1} \to H^{2} (C, T_{C}^{ns}) = 0,$ where $T_{C}^{ns}$ is the tangent sheaf of the normalisation pulled back and the local $T_{p_{i}}^{1} ≅ k$ is the one-dimensional local-node-smoothing direction at $p_{i}$ . The dimension is $dim H^{1} (C, T_{C}^{ns}) + δ$ , where the first term is the deformation dimension of the normalisation and $δ$ is the number of nodes. For a stable nodal curve of arithmetic genus $g$ with $δ$ nodes, this dimension equals $3 g - 3$ — matching the deformation dimension of a smooth curve of the same genus, with $δ$ node-smoothing directions traded for $δ$ moduli of the smooth components.

Deligne-Mumford compactification. The Deligne-Mumford compactification $\overline{M_{g}}$ (Deligne-Mumford 1969 ^{[Deligne-Mumford 1969]}) is the moduli space of stable curves of genus $g$ : connected projective curves of arithmetic genus $g$ with only nodal singularities and finite automorphism group. The local geometry of $\overline{M_{g}}$ at a nodal curve $[C]$ is governed by the deformation theory above: a versal deformation of $C$ has $3 g - 3$ parameters, of which $δ$ are node-smoothing parameters. The boundary divisors $Δ_{i}$ of $\overline{M_{g}}$ correspond to the loci where specific nodes do not smooth (the $t_{i} = 0$ locus in the versal-deformation base), and the universal family $\overline{C_{g}} \to \overline{M_{g}}$ is the family of stable curves obtained by glueing local versal deformations. Smoothness of $\overline{M_{g}}$ at $[C]$ (or smoothness of the Deligne-Mumford stack at $[C]$ ) follows from the unobstructedness of the deformation functor, which extends to nodal curves because $Ext^{2} (Ω_{C}^{1}, O_{C}) = 0$ for any reduced curve $C$ (by a dimension-one Ext vanishing argument). Rubric: full credit for the node-smoothing description, the dimension formula $3 g - 3 = dim H^{1} (C, T_{C}^{ns}) + δ$ , the role in $\overline{M_{g}}$ , and the unobstructedness argument.

Advanced results Master

Theorem (Kuranishi 1962 — existence of a miniversal family in the analytic category). Let $X$ be a compact complex manifold. There exists a complete miniversal deformation $X \to (B, 0)$ of $X$ over a germ of an analytic space $(B, 0)$ , characterised by the property that the Kodaira-Spencer map $ρ_{0} : T_{0} B \to H^{1} (X, T_{X})$ is an isomorphism. The germ $(B, 0)$ is smooth iff $H^{2} (X, T_{X}) = 0$ ; in general $(B, 0)$ is a closed analytic subspace of $H^{1} (X, T_{X})$ cut out by holomorphic equations whose linear part vanishes and whose obstruction quadratic part takes values in $H^{2} (X, T_{X})$ . ^{[Kuranishi 1962]}

For a smooth projective curve $C$ the obstruction space $H^{2} (C, T_{C}) = 0$ vanishes, so the Kuranishi family is smooth — the miniversal deformation base is an open neighbourhood of the origin in $H^{1} (C, T_{C}) ≅ C^{3 g - 3}$ , and the family extends to a formal family over the power series ring $C [[t_{1}, \dots, t_{3 g - 3}]]$ . This is the analytic counterpart of the algebraic Schlessinger pro-representability result, and the two pictures agree on the dimension count and the smoothness of the deformation space.

Theorem (Schlessinger 1968 — functor of Artin rings, full statement). Let $F : Art_{k} \to Set$ be a covariant functor with $F(k) = {} $. T h eco n d i t i o n s (H 1) an d (H 2) o f t h e p r e v i o u se x er c i se a r ee q u i v a l e n tt o t h ee x i s t e n ceo f ah u l l — amini v er s a l p r o - r e p r ese n t in g co m pl e t e l oc a l N oe t h er ian$ k $- a l g e b r a$ R $t o g e t h er w i t ha s m oo t h s u r j ec t i o n$ h_R \to F $o n p o in t so f A r t in r in g s, in d u c in g ani so m or p hi s m o n t an g e n t s p a ces$ \mathrm{Hom}(\mathfrak{m}_R / \mathfrak{m}_R^2, k) \cong t_F $. T h e f u l l co n d i t i o n s (H 1) - (H 4) h o l d i f f$ F $i s p r o - r e p r ese n t e d b y$ R$.* ^{[Schlessinger 1968]}

Schlessinger's criterion is the algebraic foundation of formal deformation theory: every deformation functor satisfying the four conditions admits a miniversal base. For curve deformations the criterion holds and the miniversal base is $Spf k [[t_{1}, \dots, t_{3 g - 3}]]$ . The proof of pro-representability uses Schlessinger's construction of a tower ${R_{n}}$ of finite-length Artin rings, with each $R_{n}$ obtained from $R_{n - 1}$ by an explicit central extension, and the inverse limit $R = lim R_{n}$ pro-representing the functor.

Theorem (Deligne-Mumford 1969 — irreducibility and the universal family of stable curves). Let $g \geq 2$ . The moduli stack $\overline{M_{g}}^{stack}$ of stable curves of genus $g$ is a proper smooth Deligne-Mumford stack over $Z$ of relative dimension $3 g - 3$ . The coarse moduli space $\overline{M_{g}}$ is a projective scheme over $Z$ . There is a universal family of stable curves $\overline{C_{g}} \to \overline{M_{g}}^{stack}$ on the stack, and the fibre over a closed point $[C]$ is the curve $C$ itself. ^{[Deligne-Mumford 1969]}

The universal family $\overline{C_{g}} \to \overline{M_{g}}^{stack}$ exists only on the stack, not on the coarse moduli space, because curves with extra (non-identity) automorphisms produce twisted families that do not descend to families on the scheme. The local structure of $\overline{C_{g}} \to \overline{M_{g}}^{stack}$ at a stable nodal curve is given by the local versal deformation of the nodal curve, glued together by the Schlessinger pro-representability. Smoothness of the stack follows from unobstructedness of nodal-curve deformations ( $Ext^{2} (Ω_{C}^{1}, O_{C}) = 0$ for any reduced curve).

Theorem (Mumford 1965 — moduli of curves as a quasi-projective scheme). For $g \geq 2$ , the coarse moduli space $M_{g}$ of smooth projective curves of genus $g$ is a quasi-projective scheme over $Z$ of dimension $3 g - 3$ , smooth at every point $[C]$ with $Aut (C) = {1}$ , and with quotient singularities at points with extra (non-identity) finite automorphism groups. ^{[Mumford 1965]}

The construction is via tri-canonical embedding + Hilbert scheme + GIT quotient (see 04.10.01). The deformation-theoretic picture refines this by providing local coordinates on $M_{g}$ at $[C]$ : the Schlessinger miniversal base $Spf k [[t_{1}, \dots, t_{3 g - 3}]]$ is the formal completion of $M_{g}$ at $[C]$ , and the global GIT quotient assembles these formal pictures into a single scheme. The matching of $dim Hilb^{P} - dim PGL$ with $3 g - 3 = dim H^{1} (C, T_{C})$ is the central reconciliation between the two constructions.

Theorem (Murre tangent-obstruction calculus). For any smooth proper morphism $π : X \to B$ between smooth Noetherian schemes, there is a Kodaira-Spencer map $\rho : T_B \to R^1 \pi_ T_{\mathcal{X}/B} $r e a l i s in g t h er e l a t i v e t an g e n t s h e a f co h o m o l o g y a s t h e u ni v er s a l so u r ceo f f i r s t - or d er d e f or ma t i o n s . W h e n$ R^2 \pi_* T_{\mathcal{X}/B} = 0 $o n$ B $(f or in s t an ce w h e n$ \pi $ha so n e - d im e n s i o na l f ib r es), e v er y d e f or ma t i o n p r o b l e m f or$ \pi $i s u n o b s t r u c t e d an d t h e f ami l y$ \mathcal{X} \to B $i s l oc a l l y v er s a l a t e v er y p o in tw h er e$ \rho$ is an isomorphism.*

Murre's tangent-obstruction formalism axiomatises the deformation theory in terms of two functorial sheaves on the base: the tangent sheaf $T^{1} = R^{1} π_{*} T_{X / B}$ (first-order deformation classifier) and the obstruction sheaf $T^{2} = R^{2} π_{*} T_{X / B}$ (obstruction classifier). For curve families $T^{2}$ vanishes identically, and the deformation theory reduces to a single first-order data sheaf $T^{1}$ of rank $3 g - 3$ . This formalism is what generalises to the moduli theory of higher-dimensional varieties: the Hilbert-scheme tangent space at $[X]$ is $H^{0} (X, N_{X / P^{n}})$ (sections of the normal bundle of $X$ in the ambient projective space), and the obstruction lies in $H^{1} (X, N_{X / P^{n}})$ , with both spaces controlled by the same long exact sequence machinery as Murre's calculus.

Theorem (Artin 1969 — algebraic approximation of formal deformations). Every formal deformation of a smooth projective curve $C$ over a complete local Noetherian $k$ -algebra $R$ extends to an algebraic deformation over a finite-type $k$ -algebra $A$ whose completion contains $R$ . In particular, the miniversal formal family of $C$ extends to a miniversal algebraic family parametrised by an étale neighbourhood of $[C]$ in $M_{g}$ .

Artin's approximation theorem (Artin 1969 Algebraic approximation of structures over complete local rings) bridges the formal-deformation theory of Kodaira-Spencer-Schlessinger with the algebraic-moduli theory of Mumford. The formal pro-representing object $Spf k [[t_{1}, \dots, t_{3 g - 3}]]$ for $Def_{C}$ is the formal completion of an étale neighbourhood of $[C]$ in $M_{g}$ , and the algebraic moduli space recovers the formal deformation theory through this completion. Together, Schlessinger and Artin give the algebraic side of what Kodaira-Spencer-Kuranishi give analytically.

Synthesis. The deformation theory of smooth projective curves identifies the tangent space to the moduli space $M_{g}$ at a curve $[C]$ with the cohomology group $H^{1} (C, T_{C})$ , and the central insight is that the obstruction space $H^{2} (C, T_{C})$ vanishes by dimension reasons because $T_{C}$ is a line bundle on a one-dimensional variety — putting these together, the deformation functor of any smooth projective curve is unobstructed and pro-represented by a power series ring in $3 g - 3$ variables, and the moduli space $M_{g}$ is smooth at every curve with identity automorphism group. Three apparently distinct constructions — the Čech-cocycle realisation of first-order deformations as glueing data, the Kodaira-Spencer connecting map for a family of curves, and the Schlessinger pro-representability criterion on Artin rings — fit together as one identity: each computes the same tangent space $H^{1} (C, T_{C})$ , and the same vanishing $H^{2} (C, T_{C}) = 0$ guarantees unobstructedness in all three frameworks. The foundational reason this works is that the tangent sheaf of a smooth curve is a line bundle of negative degree for $g \geq 2$ , and Riemann-Roch on this line bundle reads off the dimension $3 g - 3$ on the nose, while dimension-one vanishing kills all higher cohomology.

This pattern appears again in 04.10.01 (moduli of curves), where the tangent space to $M_{g}$ at $[C]$ is identified with $H^{1} (C, T_{C})$ and the Mumford-Schlessinger reconciliation of GIT-quotient dimension with deformation-theoretic tangent dimension is the bridge between the global and local constructions, and builds toward the universal family of stable curves $\overline{C_{g}} \to \overline{M_{g}}^{stack}$ of Deligne-Mumford, where the deformation theory of nodal curves controls the boundary structure. The pattern recurs in higher dimension: for any smooth projective variety $X$ , the tangent space to the local Kuranishi space is $H^{1} (X, T_{X})$ and the obstruction space is $H^{2} (X, T_{X})$ , but for $dim X \geq 2$ the obstruction space can be non-zero and the local moduli space can be singular. The deeper structural fact is that the cotangent complex $L_{X}$ of Illusie 1971 promotes this picture to the derived setting, with the deformation tangent space realised as $Ext^{1} (L_{X}, O_{X})$ and the obstruction as $Ext^{2} (L_{X}, O_{X})$ , recovering the curve theory as the dimension-one case where the cotangent complex is concentrated in degree zero and the Ext-vanishing follows from dimension. The bridge is that the Kodaira-Spencer programme of 1958, the Schlessinger pro-representability of 1968, the Mumford GIT construction of 1965, and the Illusie cotangent complex of 1971 all converge on the same identification of $H^{1} (C, T_{C})$ as the universal first-order deformation classifier of a smooth projective curve.

The synthesis is structural: every classical moduli construction for curves — Riemann's $3 g - 3$ , Mumford's GIT quotient, Deligne-Mumford's stable-curve compactification, Schlessinger's pro-representable hull — reduces to the cohomological count of $H^{1} (C, T_{C}) = 3 g - 3$ and the vanishing $H^{2} (C, T_{C}) = 0$ . The deformation theory of smooth curves is what makes the moduli space well-behaved, and the vanishing of the obstruction is what makes it smooth.

Full proof set Master

Proposition (dimension of $H^{1} (C, T_{C})$ via Riemann-Roch). Let $C$ be a smooth projective curve of genus $g$ over an algebraically closed field $k$ . Then $dim_{k} H^{1} (C, T_{C}) = 3 g - 3$ for $g \geq 2$ , $dim_{k} H^{1} (C, T_{C}) = 1$ for $g = 1$ , and $dim_{k} H^{1} (C, T_{C}) = 0$ for $g = 0$ .

Proof. The tangent line bundle $T_{C} = Ω_{C}^{1}^{\lor}$ has degree $de g T_{C} = - de g Ω_{C}^{1} = - (2 g - 2) = 2 - 2 g$ .

Case $g = 0$ : $C = P^{1}$ and $T_{P^{1}} = O_{P^{1}} (2)$ has degree 2. By the cohomology of line bundles on $P^{1}$ (see 04.03.04), $h^{0} (O_{P^{1}} (2)) = 3$ and $h^{1} (O_{P^{1}} (2)) = 0$ . So $dim H^{1} (P^{1}, T_{P^{1}}) = 0$ , matching the formula $3 \cdot 0 - 3 = - 3$ corrected to zero (the formula $3 g - 3$ is valid for $g \geq 2$ ; for $g \leq 1$ the dimension is $max (0, 3 g - 3 + dim H^{0} (C, T_{C}))$ , which gives 0 at $g = 0$ and 1 at $g = 1$ ).

Case $g = 1$ : $T_{C} = O_{C}$ (the canonical bundle of a genus-1 curve is the structure sheaf, and the tangent bundle is its dual). Then $h^{0} (O_{C}) = 1$ and $h^{1} (O_{C}) = g = 1$ . So $dim H^{1} (C, T_{C}) = 1$ , matching the modulus count for elliptic curves.

Case $g \geq 2$ : $de g T_{C} = 2 - 2 g < 0$ , and a line bundle of negative degree on a smooth projective curve has no non-zero global sections (otherwise the section would define an effective divisor of non-negative degree, contradicting the negative degree of the line bundle's first Chern class). So $h^{0} (C, T_{C}) = 0$ . Riemann-Roch for the line bundle $T_{C}$ on $C$ reads (see 04.04.01) $$ h^0(C, T_C) - h^1(C, T_C) = \deg T_C + 1 - g = (2 - 2g) + 1 - g = 3 - 3g. $$ Substituting $h^{0} (C, T_{C}) = 0$ gives $h^{1} (C, T_{C}) = 3 g - 3$ . $□$

Proposition ( $H^{2} (C, T_{C}) = 0$ for any smooth projective curve $C$ ). Let $C$ be a smooth projective curve over an algebraically closed field $k$ . Then $H^{i} (C, F) = 0$ for every coherent sheaf $F$ on $C$ and every $i \geq 2$ . In particular $H^{2} (C, T_{C}) = 0$ .

Proof. $C$ is a one-dimensional Noetherian scheme. By Grothendieck's vanishing theorem ^{[source pending]} (Hartshorne III.2.7), $H^{i} (X, F) = 0$ for every Noetherian topological space $X$ of dimension $\leq n$ and every $i > n$ and every abelian sheaf $F$ . For $C$ with $dim C = 1$ and $i = 2$ , the theorem gives $H^{2} (C, F) = 0$ for every abelian sheaf $F$ , in particular for every coherent sheaf and in particular for $T_{C}$ . $□$

Equivalently: $C$ has an affine open cover with two open sets $U_{0}, U_{1}$ (any smooth projective curve over $k$ admits a cover by two affine opens, for instance complement-of-a-point opens of a tri-canonical embedding into projective space). The Čech complex of any sheaf $F$ on this two-element cover has only the $\overset{ˇ}{C}^{0} \to \overset{ˇ}{C}^{1}$ differential and vanishes in degree $\geq 2$ , so $\overset{ˇ}{H}^{i} (U, F) = 0$ for $i \geq 2$ , and this Čech cohomology agrees with sheaf cohomology by Leray's theorem (the affine opens have vanishing higher cohomology for quasi-coherent sheaves).

Proposition (first-order deformations as $H^{1} (C, T_{C})$ , full proof). The bijection $Def_{C} (k [ϵ] / ϵ^{2}) ≅ H^{1} (C, T_{C})$ established in the Intermediate-tier theorem is a $k$ -vector space isomorphism, where the addition on $Def_{C} (k [ϵ] / ϵ^{2})$ is induced by the addition of Čech cocycles on the matching open cover.

Proof. The construction of the bijection (Čech cocycle from an open cover + glueing data) is given in the Intermediate proof. The $k$ -vector space structure: a first-order deformation $C_{η}$ classified by the cohomology class $η = [(θ_{ij})]$ , scaled by $λ \in k$ , gives the deformation classified by $λ \cdot η = [(λ \cdot θ_{ij})]$ , where $λ \cdot θ_{ij}$ is the scalar multiple of the derivation $θ_{ij} \in T_{C} (U_{ij})$ . The sum of two cohomology classes $η + η^{'} = [(θ_{ij} + θ_{ij}^{'})]$ corresponds to the deformation whose glueing data is the sum of the two glueing data, modulo the equivalence introduced by changes of local trivialisation (which act by Čech coboundaries on the cocycles). The bijection respects both operations, so it is a $k$ -linear isomorphism. $□$

Theorem (unobstructedness of $Def_{C}$ for smooth projective curve). The deformation functor $Def_{C}$ of a smooth projective curve $C$ is unobstructed: every first-order deformation lifts to a formal family over $Spf k [[t_{1}, \dots, t_{3 g - 3}]]$ for $g \geq 2$ , over $Spf k [[t]]$ for $g = 1$ , and is unique (the constant family) for $g = 0$ .

Proof. Induct on the order of the lift. The base case is the first-order deformation $C_{1}$ classified by $η \in H^{1} (C, T_{C})$ . For the inductive step, suppose $C_{n}$ is a deformation over $A_{n} = k [ϵ] / ϵ^{n + 1}$ and consider lifting to $A_{n + 1} = k [ϵ] / ϵ^{n + 2}$ via the small extension $0 \to k \to A_{n + 1} \to A_{n} \to 0$ . The obstruction is a class $obs (C_{n}) \in H^{2} (C, T_{C}) \otimes_{k} k = H^{2} (C, T_{C}) = 0$ by the previous proposition. Therefore $C_{n}$ lifts to $A_{n + 1}$ , and by induction the full formal family over $k [[ϵ]]$ exists. Passing from one-parameter to $(3 g - 3)$ -parameter lifts is the analogous induction over a tower of small extensions of the higher-dimensional Artin rings $k [[t_{1}, \dots, t_{3 g - 3}]] / m^{N}$ for $N \to \infty$ ; the same $H^{2} (C, T_{C}) = 0$ vanishing kills every obstruction. By Schlessinger pro-representability (or by direct construction), the inverse limit family exists over $Spf k [[t_{1}, \dots, t_{3 g - 3}]]$ . $□$

Theorem (Kodaira-Spencer isomorphism for a versal family), stated without proof here — full proof in Kodaira-Spencer 1958 ^{[Kodaira-Spencer 1958]} and Sernesi §3.3. For the miniversal deformation $C \to (B, 0)$ of a smooth projective curve $C$ , the Kodaira-Spencer map $ρ_{0} : T_{0} B \to H^{1} (C, T_{C})$ is an isomorphism. Conversely, any deformation $C^{'} \to (B^{'}, 0^{'})$ whose Kodaira-Spencer map at $0^{'}$ is an isomorphism is miniversal. The proof is the analytic counterpart of the Schlessinger algebraic argument: the local product structure of the family at higher orders is controlled by the obstruction space $H^{2} (C, T_{C}) = 0$ , and the Kodaira-Spencer map records the first-order obstruction-free deformation directions. The miniversality follows from bijectivity of $ρ_{0}$ on the tangent space level and unobstructedness at higher order.

Theorem (Schlessinger 1968 pro-representability of $Def_{C}$ ), stated without proof here — full proof in Schlessinger 1968 ^{[Schlessinger 1968]} §2. For $C$ a smooth projective curve with $Aut (C) = {1}$ , the deformation functor $Def_{C}$ satisfies the four Schlessinger conditions (H1)-(H4) and is pro-represented by the complete local Noetherian ring $k [[t_{1}, \dots, t_{3 g - 3}]]$ for $g \geq 2$ . For curves with extra (non-identity) automorphism group, conditions (H1)-(H3) hold but (H4) fails, and $Def_{C}$ admits only a hull, not a pro-representing object. The construction of the pro-representing tower is the Schlessinger inductive limit of finite-length Artin algebras $R_{n}$ , each producing one more order of lift, with the inverse limit $R = lim R_{n}$ providing the universal pro-representing object.

Connections Master

Sheaf of differentials 04.08.01. The deformation theory of a smooth scheme $X$ is controlled by the cohomology of the tangent sheaf $T_{X} = Ω_{X}^{1}^{\lor}$ , the dual of the cotangent sheaf. For curves, the tangent sheaf is a line bundle and Riemann-Roch gives an exact dimension count. The conormal-sequence machinery used to compute $Ω_{X}^{1}$ for embedded curves [see 04.05.07] feeds directly into the cohomological computation of $H^{1} (C, T_{C})$ via Serre duality.
Canonical sheaf 04.08.02. The tangent line bundle $T_{C}$ is dual to the canonical sheaf $ω_{C} = Ω_{C}^{1}$ . The negative-degree characterisation of $T_{C}$ for $g \geq 2$ — namely $de g T_{C} = 2 - 2 g < 0$ — comes from the degree-genus identity $de g ω_{C} = 2 g - 2$ that defines the canonical class. The vanishing $h^{0} (T_{C}) = 0$ for $g \geq 2$ is the negative-degree side of the global-sections-or-vanishing dichotomy for line bundles on curves.
Riemann-Roch for curves 04.04.01. The dimension count $dim H^{1} (C, T_{C}) = 3 g - 3$ uses Riemann-Roch on the tangent line bundle of negative degree. Riemann-Roch is the universal tool for computing cohomology dimensions of line bundles on curves, and the deformation theory of curves is its first major moduli-theoretic application.
Moduli of curves 04.10.01. The deformation theory provides the local-at-a-point picture of $M_{g}$ : the tangent space to $M_{g}$ at a curve $[C]$ with identity automorphism group is $H^{1} (C, T_{C})$ , and the local structure of $M_{g}$ is given by the Schlessinger miniversal family over $Spf k [[t_{1}, \dots, t_{3 g - 3}]]$ . The global GIT construction of Mumford 1965 assembles these formal pictures into a quasi-projective scheme, and the matching of $3 g - 3$ on both sides is the bridge.
Sheaf cohomology 04.03.01. The deformation classifier $H^{1} (C, T_{C})$ and the obstruction space $H^{2} (C, T_{C})$ are both sheaf cohomology groups. The Čech-cocycle realisation of first-order deformations identifies the tangent sheaf cohomology with glueing data of infinitesimal families, and the dimension-one vanishing of higher cohomology is what makes curve deformations unobstructed. Sheaf cohomology is the universal tool of modern deformation theory.
Smooth, étale, and unramified morphisms 04.02.05. The deformation theory of a smooth scheme $X$ uses the smoothness hypothesis at every step: smoothness of the affine open sets makes the local first-order deformations identity, smoothness of $X$ globally makes the tangent sheaf locally free, and smoothness of the projection $X \to B$ in a family makes the Kodaira-Spencer connecting sequence exact. Without smoothness, the deformation theory becomes singular (cotangent-complex-valued rather than tangent-sheaf-valued).
Geometric invariant theory 04.10.02. Mumford's GIT construction of $M_{g}$ via tri-canonical embedding and quotient by $PGL$ produces the global moduli space whose local structure is governed by the deformation theory developed here. The matching of $dim Hilb^{P} - dim PGL$ on the GIT side with $dim H^{1} (C, T_{C}) = 3 g - 3$ on the deformation-theory side is the bridge between the global and local constructions.

Historical & philosophical context Master

The deformation theory of complex analytic manifolds was founded by Kunihiko Kodaira and Donald Spencer in their two-part paper On deformations of complex analytic structures (Annals of Mathematics 67, 1958, parts I and II) ^{[Kodaira-Spencer 1958]}. Kodaira and Spencer introduced the deformation functor of a compact complex manifold $X$ , identified the first-order deformations with $H^{1} (X, T_{X})$ via a Čech-cocycle realisation, and established the Kodaira-Spencer map $ρ : T_{b} B \to H^{1} (X_{b}, T_{X_{b}})$ for a holomorphic family $π : X \to B$ . They proved the Kodaira-Spencer completeness theorem: a family is locally complete (miniversal) at a point iff the Kodaira-Spencer map is an isomorphism at that point. The Annals paper was the first systematic deformation theory of complex structures, and it set the framework that all subsequent algebraic deformation theory has followed.

Masatake Kuranishi's 1962 On the locally complete families of complex analytic structures (Annals of Mathematics 75, 1962) ^{[Kuranishi 1962]} proved the existence of a miniversal local family for every compact complex manifold, with the deformation base cut out from $H^{1} (X, T_{X})$ by holomorphic equations whose obstruction quadratic part lies in $H^{2} (X, T_{X})$ . For a smooth projective curve $C$ the obstruction space $H^{2} (C, T_{C})$ vanishes by dimension reasons, and the Kuranishi family is smooth of dimension $3 g - 3$ . Kuranishi's theorem is the analytic counterpart of the algebraic Schlessinger pro-representability of 1968, and the two together establish the existence and structure of the universal local deformation in both the analytic and algebraic categories.

The algebraic deformation theory was systematised by Michael Schlessinger's 1968 Functors of Artin rings (Trans. AMS 130, 208-222) ^{[Schlessinger 1968]}, where Schlessinger introduced the four conditions (H1)-(H4) on a functor $F : Art_{k} \to Set$ and proved that (H1)-(H3) imply existence of a miniversal hull, while (H1)-(H4) imply full pro-representability. For curve deformations the four conditions hold (with (H4) requiring the identity-automorphism hypothesis), and $Def_{C}$ is pro-represented by $k [[t_{1}, \dots, t_{3 g - 3}]]$ . David Mumford's 1965 Geometric Invariant Theory (Springer) ^{[Mumford 1965]} and his 1966 Lectures on Curves on an Algebraic Surface (Princeton Annals 59) constructed the global moduli space $M_{g}$ as a quasi-projective scheme of dimension $3 g - 3$ , with the deformation theory of Kodaira-Spencer-Kuranishi-Schlessinger providing the local-at-a-point picture and the Mumford GIT quotient providing the global scheme structure. The reconciliation of the two pictures — that $dim H^{1} (C, T_{C}) = 3 g - 3 = dim Hilb^{P} - dim PGL$ — is the central numerical identity of the moduli theory of curves.

Pierre Deligne and David Mumford's 1969 The irreducibility of the space of curves of given genus (Publ. Math. IHES 36, 75-109) ^{[Deligne-Mumford 1969]} extended the moduli theory to the compactification $\overline{M_{g}}$ via stable curves, and introduced the stack-theoretic framework that handles curves with extra (non-identity) automorphism groups. The deformation theory of nodal curves entering the boundary of $\overline{M_{g}}$ uses the same $H^{1}$ and $H^{2}$ machinery, with the local versal deformation of each node providing one extra modulus direction. The full development of the deformation theory of curves in textbook form is given in Joe Harris and Ian Morrison's Moduli of Curves (Springer GTM 187, 1998) ^{[Harris-Morrison 1998]}, with the modern algebraic synthesis in Edoardo Sernesi's Deformations of Algebraic Schemes (Springer Grundlehren 334, 2006) ^{[Sernesi 2006]}. The Sernesi treatment integrates the Schlessinger framework with Luc Illusie's 1971 cotangent complex (which generalises the tangent-sheaf cohomology to higher-derived information), and applies the resulting machinery to the deformation theory of curves, surfaces, threefolds, and singular schemes.

Bibliography Master

@book{HarrisMorrison1998,
  author    = {Harris, Joe and Morrison, Ian},
  title     = {Moduli of Curves},
  publisher = {Springer-Verlag},
  series    = {Graduate Texts in Mathematics},
  volume    = {187},
  year      = {1998}
}

@book{Sernesi2006,
  author    = {Sernesi, Edoardo},
  title     = {Deformations of Algebraic Schemes},
  publisher = {Springer-Verlag},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {334},
  year      = {2006}
}

@article{KodairaSpencer1958I,
  author  = {Kodaira, Kunihiko and Spencer, Donald C.},
  title   = {On deformations of complex analytic structures, {I}, {II}},
  journal = {Annals of Mathematics},
  volume  = {67},
  year    = {1958},
  pages   = {328--401 and 403--466}
}

@article{Kuranishi1962,
  author  = {Kuranishi, Masatake},
  title   = {On the locally complete families of complex analytic structures},
  journal = {Annals of Mathematics},
  volume  = {75},
  year    = {1962},
  pages   = {536--577}
}

@article{Schlessinger1968,
  author  = {Schlessinger, Michael},
  title   = {Functors of {A}rtin rings},
  journal = {Transactions of the American Mathematical Society},
  volume  = {130},
  year    = {1968},
  pages   = {208--222}
}

@book{Mumford1965GIT,
  author    = {Mumford, David},
  title     = {Geometric Invariant Theory},
  publisher = {Springer-Verlag},
  year      = {1965},
  note      = {3rd ed. with Fogarty and Kirwan, 1994}
}

@book{Mumford1966Curves,
  author    = {Mumford, David},
  title     = {Lectures on Curves on an Algebraic Surface},
  publisher = {Princeton University Press},
  series    = {Annals of Mathematics Studies},
  volume    = {59},
  year      = {1966}
}

@article{DeligneMumford1969,
  author  = {Deligne, Pierre and Mumford, David},
  title   = {The irreducibility of the space of curves of given genus},
  journal = {Publications Math{\'e}matiques de l'IH{\'E}S},
  volume  = {36},
  year    = {1969},
  pages   = {75--109}
}

@book{ArbarelloCornalbaGriffiths2011,
  author    = {Arbarello, Enrico and Cornalba, Maurizio and Griffiths, Phillip A.},
  title     = {Geometry of Algebraic Curves, Volume {II}},
  publisher = {Springer-Verlag},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {268},
  year      = {2011}
}

@article{Artin1969,
  author  = {Artin, Michael},
  title   = {Algebraic approximation of structures over complete local rings},
  journal = {Publications Math{\'e}matiques de l'IH{\'E}S},
  volume  = {36},
  year    = {1969},
  pages   = {23--58}
}

@article{Illusie1971,
  author  = {Illusie, Luc},
  title   = {Complexe cotangent et d{\'e}formations {I}, {II}},
  journal = {Lecture Notes in Mathematics},
  volume  = {239 and 283},
  year    = {1971--1972},
  publisher = {Springer-Verlag}
}

@article{Riemann1857,
  author  = {Riemann, Bernhard},
  title   = {Theorie der {A}belschen {F}unctionen},
  journal = {Journal f{\"u}r die reine und angewandte Mathematik},
  volume  = {54},
  year    = {1857},
  pages   = {115--155}
}

Prerequisites

04.08.01
04.08.02
04.04.01
04.10.01
04.03.01

Tier anchors

beginner: Strogatz-style: a smooth curve admits a family of small wiggles, and the count of independent wiggles is $3g - 3$
intermediate: Harris-Morrison *Moduli of Curves* Ch. 1-2; Sernesi *Deformations of Algebraic Schemes* §1.2-§1.3; Hartshorne *Deformation Theory* §1; Vakil *FOAG* Ch. 25
master: Harris-Morrison *Moduli of Curves* (Springer GTM 187, 1998); Sernesi *Deformations of Algebraic Schemes* (Springer Grundlehren 334, 2006); Kodaira-Spencer 1958 *On deformations of complex analytic structures I, II* (Ann. of Math. 67); Kuranishi 1962 *On the locally complete families of complex analytic structures* (Ann. of Math. 75); Schlessinger 1968 *Functors of Artin rings* (Trans. AMS 130); Mumford 1965 *Geometric Invariant Theory*; Deligne-Mumford 1969 *The irreducibility of the space of curves of given genus* (Publ. Math. IHES 36)

References

Harris-Morrison — Moduli of Curves · Springer GTM 187, 1998; Chapter 1 (deformation theory of smooth curves), Chapter 2 (stable curves and the Deligne-Mumford compactification)
Sernesi — Deformations of Algebraic Schemes · Springer Grundlehren 334, 2006; §1.2 (first-order deformations and $H^1(T_X)$), §2.4 (obstructions and $H^2(T_X)$), §3.3 (deformations of curves and the Kodaira-Spencer map)
Kodaira-Spencer — On deformations of complex analytic structures I, II · Ann. of Math. 67 (1958), 328-401 and 403-466
Kuranishi — On the locally complete families of complex analytic structures · Ann. of Math. 75 (1962), 536-577
Schlessinger — Functors of Artin rings · Trans. Amer. Math. Soc. 130 (1968), 208-222
Mumford — Geometric Invariant Theory · Springer 1965; 3rd ed. Mumford-Fogarty-Kirwan 1994; §5 (construction of $\mathcal{M}_g$ via tri-canonical embedding)
Mumford — Lectures on Curves on an Algebraic Surface · Princeton Annals of Mathematics Studies 59, 1966; Lectures 22-27 on deformations
Deligne-Mumford — The irreducibility of the space of curves of given genus · Publ. Math. IHES 36 (1969), 75-109; stable curves and the compactification $\overline{\mathcal{M}_g}$
Arbarello-Cornalba-Griffiths — Geometry of Algebraic Curves Volume II · Springer Grundlehren 268, 2011; Ch. XI (deformations of nodal curves and the universal family of stable curves)

Estimated time

beginner: 15m
intermediate: 45m
master: 80m