Schottky problem

Intuition [Beginner]

Take a compact Riemann surface $X$ — a closed surface dressed with the data needed to ask whether a function on it is holomorphic — of genus $g$. Riemann's bilinear relations attach to $X$ a $g\times g$ symmetric complex matrix $\tau$ whose imaginary part is positive definite, an element of the Siegel upper half-space. The matrix $\tau$ packages the entire complex-analytic fingerprint of $X$. Reversing the construction, every such matrix $\tau$ defines its own complex torus that behaves like the Jacobian of some hypothetical curve.

So which complex tori actually come from curves? The Schottky problem asks for the boundary between Jacobians of real Riemann surfaces and tori that merely look like Jacobians. The dimension count says it must be drawn: the moduli space of curves of genus $g$ has dimension $3g-3$, but the space of candidate matrices has dimension $g(g+1)/2$. For genus $g\le 3$ the two dimensions match, and every candidate is a Jacobian. For genus $g\ge 4$ the curves form a strict subvariety of the candidate space, and the Schottky problem asks for the equations that cut out this subvariety.

Why bother? Because the answer ties together algebraic geometry, the theory of theta functions, and a remarkable family of integrable wave equations from soliton physics. Schottky's 1888 polynomial relation, Novikov's KP-hierarchy conjecture (proved by Shiota in 1986), and the Welters trisecant criterion (proved by Krichever in 2010) all give different but equivalent answers — and each one builds a bridge between the geometry of curves and another corner of mathematics.

Visual [Beginner]

A schematic shows the Siegel upper half-space ${\mathfrak{H}}_g$ as a large blob labelled "all candidate period matrices, dimension $g(g+1)/2$". Inside the blob a smaller closed region is labelled "Jacobi locus ${\mathcal{J}}_g$, dimension $3g-3$" and is shaded; for $g=4$ the smaller region is a hypersurface cut out by Schottky's polynomial $S(\theta )=0$, indicated by a single explicit equation drawn beside it. An arrow from the moduli space ${\mathcal{M}}_g$ of curves into the Jacobi locus traces the period mapping; a second arrow indicates the conjectured Andreotti-Mayer locus extending around the shaded region.

Worked example [Beginner]

Take genus $g=3$. The candidate space ${\mathfrak{H}}_3$ of symmetric $3\times 3$ complex matrices with positive-definite imaginary part has complex dimension $3\cdot 4/2=6$. The moduli space of genus-$3$ curves has dimension $3\cdot 3-3=6$. The two dimensions match, so the Schottky problem in genus $3$ has a vacuous answer: every principally polarised abelian variety of dimension $3$ is the Jacobian of some genus-$3$ curve (or a product of lower-genus Jacobians on a thin boundary). No equations are needed — the period map is a birational equivalence.

For $g=4$ the dimension count flips. The candidate space ${\mathfrak{H}}_4$ has complex dimension $4\cdot 5/2=10$. The moduli space of genus-$4$ curves has dimension $3\cdot 4-3=9$. The Jacobi locus is a codimension-$1$ subvariety, a hypersurface inside ${\mathfrak{H}}_4/\mathrm{Sp}(8,\mathbb{Z})$. Schottky 1888 gave a single polynomial equation $S(\theta )=0$ in theta-constants that cuts out this hypersurface. So in genus $4$ the Schottky problem has a clean explicit solution; for $g\ge 5$ the codimension grows quickly to $(g-2)(g-3)/2$, and explicit polynomial equations are not known.

What this tells us: dimension counting is the first cut. For $g\le 3$ there is nothing to ask — the candidates and the Jacobians agree. The real problem starts at $g=4$, where Schottky's quartic gave the first explicit answer, and for $g\ge 5$ the answer comes from integrable systems (the KP hierarchy) or projective geometry (the trisecant criterion) rather than from a single polynomial.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $g\ge 1$. The Siegel upper half-space is $$ \mathfrak{H}_g := {\tau \in M_g(\mathbb{C}) : \tau = \tau^T,\ \mathrm{Im},\tau > 0}, $$ a connected complex domain of complex dimension $g(g+1)/2$. The symplectic group $\mathrm{Sp}(2g,\mathbb{Z})$ acts on ${\mathfrak{H}}_g$ by $\left(\begin{array}{cc}A& B\\ C& D\end{array}\right)\cdot \tau :=(A\tau +B)(C\tau +D{)}^{-1}$, and the quotient $$ \mathcal{A}_g := \mathfrak{H}_g / \mathrm{Sp}(2g, \mathbb{Z}) $$ is the coarse moduli space of principally polarised abelian varieties of complex dimension $g$ (cf. 06.06.07).

Let ${\mathcal{M}}_g$ be the moduli space of smooth projective complex curves of genus $g$, of complex dimension $3g-3$ for $g\ge 2$ (Riemann's count). The period mapping (cf. 06.08.02) $$ \mathrm{Per} : \mathcal{M}_g \to \mathcal{A}_g, \qquad X \mapsto \mathrm{Jac}(X), $$ sends a curve to its principally polarised Jacobian, and is injective by Torelli's theorem (Torelli 1913). The Jacobi locus is its image: $$ \mathcal{J}_g := \mathrm{Per}(\mathcal{M}_g) \subset \mathcal{A}_g, $$ a quasi-projective subvariety of complex dimension $3g-3$ for $g\ge 2$.

Codimension count. ${\mathrm{codim}}_{{\mathcal{A}}_g}{\mathcal{J}}_g=g(g+1)/2-(3g-3)=(g-2)(g-3)/2$. Hence ${\mathcal{J}}_g={\mathcal{A}}_g$ iff $g\le 3$, and ${\mathcal{J}}_g⊊{\mathcal{A}}_g$ is a strict subvariety for $g\ge 4$.

The Schottky problem. Find explicit conditions on $\tau \in {\mathfrak{H}}_g$ that characterise membership $[\tau ]\in {\mathcal{J}}_g$ inside ${\mathcal{A}}_g$. Equivalently, identify the period matrices of smooth projective curves among all symmetric positive-imaginary $g\times g$ complex matrices, modulo the symplectic-modular action.

Theta-constants. For a half-period characteristic $(m,m^{\prime })\in (\frac{1}{2}{\mathbb{Z}}^g/{\mathbb{Z}}^g{)}^2$, the theta-constant is $$ \thetam, m' := \sum_{n \in \mathbb{Z}^g} \exp\bigl(\pi i (n + m)^T \tau (n + m) + 2\pi i (n + m)^T m'\bigr), $$ a holomorphic function on ${\mathfrak{H}}_g$. Theta-constants $\theta [m,m^{\prime }]$ are Siegel modular forms of weight $1/2$ on the principal congruence subgroup $\Gamma (2)\subset \mathrm{Sp}(2g,\mathbb{Z})$, and the ring they generate (modulo a known kernel) is dense in the ring of all Siegel modular forms.

Counterexamples to common slips.

• The Schottky problem is not asking which complex tori are abelian varieties — that's the polarisation question, answered by the Riemann bilinear relations (cf. 06.06.07). The Schottky problem starts where bilinear relations end: every $\tau \in {\mathfrak{H}}_g$ is a principally polarised abelian variety; the question is which ones come from curves.
• The locus ${\mathcal{J}}_g$ is the closure of the open part where the curve is non-hyperelliptic-modulo-isomorphism; the boundary is the locus of products of Jacobians and the hyperelliptic Jacobi sublocus, both of which are themselves subvarieties of ${\mathcal{A}}_g$.
• The codimension formula $(g-2)(g-3)/2$ is positive only for $g\ge 4$; the cases $g=1,2,3$ have $(g-2)(g-3)/2\le 0$, encoding that every PPAV of dimension $\le 3$ is a Jacobian (or a product of Jacobians of genus $\le 1$ and $\le 2$ in the case $g=2$).
• An "explicit equation" cutting out ${\mathcal{J}}_g$ in genus $\ge 5$ is not known in the polynomial-modular-form sense; what is known is the equivalent characterisation by the KP hierarchy (Shiota 1986) and the trisecant criterion (Krichever 2010). These are honest equations — just not single polynomials in theta-constants.

Key theorem with proof [Intermediate+]

Theorem (Schottky 1888, genus-$4$ Schottky relation). There exists a polynomial $S$ in theta-constants such that for $\tau \in {\mathfrak{H}}_4$, $$ [\tau] \in \mathcal{J}_4 \iff S(\thetam, m') = 0. $$ Concretely, the locus $\{S=0\}\subset {\mathfrak{H}}_4/\mathrm{Sp}(8,\mathbb{Z})$ is an irreducible hypersurface of codimension $1$ in ${\mathcal{A}}_4$, and this hypersurface coincides with the Jacobi locus ${\mathcal{J}}_4$.

Proof (sketch, following Schottky 1888 and Igusa's modern reformulation). The argument has four steps: codimension count, construction of $S$ from the Schottky-Jung framework, vanishing of $S$ on Jacobians via Riemann's vanishing theorem, and irreducibility of the resulting hypersurface.

Step 1 — codimension count. By the dimension formula, $$ \dim_{\mathbb{C}} \mathcal{A}4 = \frac{4 \cdot 5}{2} = 10, \quad \dim{\mathbb{C}} \mathcal{M}_4 = 3 \cdot 4 - 3 = 9. $$ By Torelli's theorem the period map ${\mathcal{M}}_4\to {\mathcal{A}}_4$ is injective, so ${\mathcal{J}}_4$ has complex dimension $9$ and codimension $1$ in ${\mathcal{A}}_4$. A characterisation by the vanishing of a single non-zero Siegel modular form is dimensionally consistent.

Step 2 — construction of $S$. For half-period characteristics $(m,m^{\prime })$ with $m,m^{\prime }\in \{0,1/2{\}}^4$, group the $2^8=256$ theta-constants $\theta [m,m^{\prime }]$ by the parity of $4(m\cdot m^{\prime })\mathrm{mod}2$. The even characteristics (parity $0$) give $2^{2g-1}+2^{g-1}=136$ theta-constants for $g=4$; the odd characteristics (parity $1$) give $120$ theta-constants, all of which vanish identically on ${\mathfrak{H}}_g$. The Schottky relation is a polynomial in the $136$ even theta-constants of ${\mathfrak{H}}_4$. The explicit form, in Igusa's normalisation, is $$ S(\theta) := \theta_1 \theta_2 \theta_3 \theta_4 - \theta_5 \theta_6 \theta_7 \theta_8 - \theta_9 \theta_{10} \theta_{11} \theta_{12}, $$ a polynomial of degree $4$ in eight specific theta-constants, where the indexing is dictated by a quadric splitting of the ${\mathbb{F}}_2$-vector space of half-period characteristics.

Step 3 — vanishing on Jacobians. By Riemann's vanishing theorem (cf. 06.06.06) applied to a smooth projective curve $X$ of genus $4$, the theta-constants $\theta [m,m^{\prime }](\tau (X))$ at the period matrix $\tau (X)$ are constrained by the geometry of the canonical curve $X\subset {\mathbb{P}}^3$. The genus-$4$ canonical model is a complete intersection of a quadric and a cubic in ${\mathbb{P}}^3$; the quadric is unique up to scaling. The two rulings of the quadric (or the single ruling, in the degenerate case where the quadric is a cone) correspond to the two even theta-characteristics of degree-$3$ vanishing-divisors of theta on the Jacobian. The resulting algebraic relation among the eight theta-constants in Step 2 is exactly $S=0$. Hence ${\mathcal{J}}_4\subseteq \{S=0\}$.

Step 4 — irreducibility and equality. The hypersurface $\{S=0\}$ is irreducible in ${\mathcal{A}}_4$ because the splitting of the half-period characteristics into the eight constants of Step 2 is dictated by a specific symplectic-modular orbit, and the resulting modular form $S$ generates the corresponding isotypical component of the Siegel-modular-form ring. Both ${\mathcal{J}}_4$ and $\{S=0\}$ are irreducible of codimension $1$ in ${\mathcal{A}}_4$, with one contained in the other; hence they coincide. $\square$

The proof above sketches Schottky's 1888 construction. Igusa 1980 On the irreducibility of Schottky's divisor (J. Fac. Sci. Univ. Tokyo Sect. IA) gave the modern proof of irreducibility, closing a gap in Schottky's original argument. Mumford Tata Lectures on Theta II §IIIa develops the construction systematically using the Schottky-Jung modular framework. The four-step structure — codimension count, modular-form construction, vanishing on Jacobians via Riemann's vanishing theorem, irreducibility — is the standard organisation; van Geemen 1984 Siegel modular forms vanishing on the moduli space of curves (Inventiones) gives a different but equivalent presentation through automorphic representations.

Bridge. The genus-$4$ Schottky relation proven here is the cleanest case of the Schottky problem: a single explicit polynomial vanishing on Jacobians, with the codimension count matching exactly. For $g\ge 5$, the codimension $(g-2)(g-3)/2$ grows quadratically, and the analogous polynomial-modular-form characterisation has not been established. Two complete characterisations are known: Shiota's 1986 KP-hierarchy theorem, descending from Krichever's 1977 construction of KP solutions from algebraic curves and Novikov's 1979 conjecture; and Krichever's 2010 trisecant criterion, descending from Welters's 1984 conjecture. Both characterise Jacobians among PPAV in every genus, but neither is a polynomial relation in theta-constants. The picture in genus $4$ — Schottky's polynomial cuts out ${\mathcal{J}}_4$ — is a coincidence of low codimension; the deeper structural answer, valid for all $g$, runs through integrable systems and the projective geometry of the Kummer variety. Combined with 06.06.07 Riemann bilinear relations, 06.06.06 Jacobi inversion, and 06.08.02 variation of Hodge structure on the Jacobian, the present unit closes the period-theoretic loop: bilinear relations identify the Siegel upper half-space as the candidate space, Jacobi inversion identifies the Jacobian with ${\mathrm{Pic}}^g(X)$, the variation of Hodge structure is the holomorphic dependence of $\tau$ on the curve, and the Schottky problem identifies which $\tau$ come from curves.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib does not currently formalise the Siegel upper half-space ${\mathfrak{H}}_g$, the moduli space ${\mathcal{A}}_g$ of principally polarised abelian varieties, the period mapping ${\mathcal{M}}_g\to {\mathcal{A}}_g$, the Jacobi locus ${\mathcal{J}}_g\subset {\mathcal{A}}_g$, theta-constants as Siegel modular forms, or the Schottky relation as a polynomial in theta-constants. A proposed signature, in Lean 4 / Mathlib syntax, sketching the target statement for the genus-$4$ case:

[object Promise]

The proof depends on names that do not currently exist in Mathlib (the Siegel upper half-space as a Hermitian symmetric domain, theta-constants as Siegel modular forms, the moduli space ${\mathcal{A}}_g$ as a quotient stack, the period map from ${\mathcal{M}}_g$, and the Schottky modular form $S$ explicitly). For the Shiota / Krichever characterisations in higher genus, additional names are needed (the KP hierarchy as a system of PDEs, the Krichever construction from algebraic curves to KP solutions, the Kummer variety embedding, and the trisecant condition). Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results [Master]

The Schottky problem is the question of identifying the period matrices of compact Riemann surfaces among all symmetric positive-imaginary $g\times g$ complex matrices, modulo the action of $\mathrm{Sp}(2g,\mathbb{Z})$. Five characterisations of ${\mathcal{J}}_g\subset {\mathcal{A}}_g$ are known, varying in explicitness and generality.

Schottky's relation (1888). For $g=4$, the Jacobi locus is a hypersurface in ${\mathcal{A}}_4$ cut out by a single polynomial $S(\theta )=0$ in even theta-constants of degree $4$. The relation arises from the canonical-curve embedding of a genus-$4$ curve as a complete intersection of a unique quadric and a cubic in ${\mathbb{P}}^3$, with the quadric ruling determining a vanishing theta-null. Igusa 1980 proved irreducibility of the Schottky divisor, completing the original argument. The Schottky-Jung relations (Schottky-Jung 1909) generalise the construction to higher genus by passing to double covers: for a curve $X$ of genus $g$ and an unramified double cover $\tilde{X}\to X$ (genus $2g-1$), the Schottky-Jung framework relates theta-constants of $\tilde{X}$ to theta-constants of $X$; this gives equations for the Prym locus and, by composition, for the Jacobi locus, but the equations are not as clean as the genus-$4$ case. Donagi 1988 The Schottky problem (Theta Functions, Bowdoin 1987) surveys the modular-form approaches.

Andreotti-Mayer characterisation. Andreotti-Mayer 1967 On period relations for abelian integrals on algebraic curves (Ann. Sc. Norm. Sup. Pisa 21) defined the locus ${\mathcal{N}}_k:=\{[A,\Theta ]\in {\mathcal{A}}_g:\dim \mathrm{Sing}(\Theta )\ge k\}$ and proved ${\mathcal{J}}_g\subseteq {\mathcal{N}}_{g-4}$ for $g\ge 4$. The proof uses Riemann's singularity theorem identifying $\mathrm{Sing}(\Theta )$ with the translated Brill-Noether locus $W_{g-1}^1$, which has dimension $\ge g-4$ on a Jacobian. The reverse containment ${\mathcal{N}}_{g-4}\subseteq {\mathcal{J}}_g$ is conjectural and false in general for $g\ge 4$: Beauville-Debarre 1986 and Debarre 1992 identified additional components of ${\mathcal{N}}_{g-4}$ (the theta-null component, the Prym component, and others). The Andreotti-Mayer locus is a closed algebraic subvariety, but the Jacobi locus is only one of its irreducible components.

Novikov-Shiota characterisation (KP hierarchy). Novikov 1979 conjectured that the Jacobi locus is characterised by the equation $$ \frac{3}{4} u_{yy} = \left( u_t - \frac{3}{2} u u_x - \frac{1}{4} u_{xxx} \right)_x, \qquad u(x, y, t) = 2 \partial_x^2 \log \theta(xV + yU + tW + z_0; \tau), $$ the Kadomtsev-Petviashvili (KP) equation, for some constant vectors $V,U,W\in {\mathbb{C}}^g$ and starting point $z_0$. Shiota 1986 Characterization of Jacobian varieties in terms of soliton equations (Inventiones 83) proved Novikov's conjecture: $(A,\Theta )$ is a Jacobian iff its theta function admits a KP-flow direction. The proof builds on Krichever's 1977 Methods of algebraic geometry in the theory of nonlinear equations (Russian Math. Surveys 32), which constructs KP solutions from algebraic curves via the Baker-Akhiezer function. Shiota's converse uses the existence of a KP direction to construct the spectral curve; modern refinements (Mulase 1984, Mulase-Yamada 1995) give the Sato Grassmannian framework and the Hirota bilinear formulation.

Welters trisecant characterisation (projective geometry of the Kummer variety). Welters 1984 conjectured: $(A,\Theta )$ is a Jacobian iff the Kummer variety $K(\tau )\subset {\mathbb{P}}^{2^g-1}$, the image of $A_{\tau }$ under the second-order theta embedding, admits a trisecant line — a line meeting $K(\tau )$ at three distinct points. Krichever 2010 Characterizing Jacobians via trisecants of the Kummer variety (Ann. of Math. 172) proved the conjecture using a refinement of the Shiota argument. The Fay trisecant identity (Fay 1973 Theta Functions on Riemann Surfaces) gives the Jacobian direction: for $z_1+z_2+z_3\in \alpha ({\mathrm{Sym}}^3(X))-3\cdot \alpha (p_0)$ on a curve $X$, the second-order theta values ${\theta }^{(2)}(z_1),{\theta }^{(2)}(z_2),{\theta }^{(2)}(z_3)$ are linearly dependent. The converse is the substantive direction. Arbarello-De Concini 1987 Another proof of a conjecture of S. P. Novikov (Duke Math. J. 54) gave an algebro-geometric reformulation bypassing the Sato Grassmannian.

Modular-forms and automorphic perspectives. Theta-constants $\theta [m,m^{\prime }]$ are Siegel modular forms of weight $1/2$ on $\Gamma (2)\subset \mathrm{Sp}(2g,\mathbb{Z})$. The graded ring of Siegel modular forms is finitely generated (Igusa 1962, 1967 for $g\le 3$; Tsuyumine 1986 for $g=4$). The Schottky relation $S(\theta )$ is a Siegel modular form vanishing on ${\mathcal{J}}_4$, of weight determined by the polynomial degree. For $g\ge 5$, the analogous "Schottky-Igusa-Tsuyumine" generators of the modular-form ring are largely unknown, and the question of whether the Jacobi locus is cut out by polynomial relations among theta-constants remains open in the polynomial-modular-form sense. Manin 1990s-2000s and Donagi 1980s-1990s developed the modular-forms perspective; Grushevsky 2009 Geometry of ${\mathcal{A}}_g$ and its compactifications (Algebraic Geometry, Seattle 2005) surveys the modern state.

Generalisations.

• Prym Schottky problem. Characterise Pryms — abelian varieties associated to unramified double covers of curves — among PPAV. The Prym locus ${\mathcal{P}}_g\subset {\mathcal{A}}_{g-1}$ (Pryms of double covers of genus-$g$ curves yield $(g-1)$-dimensional PPAV) has dimension $3g-3$ and codimension $(g-2)(g-3)/2-1$ in ${\mathcal{A}}_{g-1}$. Beauville 1977 and Donagi 1981 developed the theory.
• Hyperelliptic Schottky. The hyperelliptic Jacobi sublocus of ${\mathcal{J}}_g$ has dimension $2g-1$ and codimension $g-2$ within ${\mathcal{J}}_g$. Mumford Tata Lectures on Theta II §IIIb gives the explicit theta-constant characterisation via vanishing properties of theta-functions associated to the Weierstrass branch points.
• Trigonal, tetragonal, $k$-gonal Schottky. Each $k$-gonality stratum gives a sublocus of ${\mathcal{J}}_g$. The Beauville-Schottky problem asks for explicit characterisations; partial results via the Welters trisecant criterion are known.
• $p$-adic Schottky uniformisation (Mumford 1972). Mumford 1972 An analytic construction of degenerating abelian varieties over complete rings (Compositio 24) gives a $p$-adic analogue: the Mumford curves are $p$-adic uniformised by $p$-adic Schottky groups, and the resulting period matrix lives in a $p$-adic Siegel upper half-space.

Synthesis. The Schottky problem closes the period-theoretic loop on compact Riemann surfaces. The bilinear relations (cf. 06.06.07) identify the Siegel upper half-space ${\mathfrak{H}}_g$ as the space of candidate period matrices; the Riemann theta function (cf. 06.06.05) and Riemann's vanishing theorem (cf. 06.06.06) realise every $\tau \in {\mathfrak{H}}_g$ as a principally polarised abelian variety with explicit transcendental theta divisor; the variation of Hodge structure on the Jacobian (cf. 06.08.02) makes the period matrix vary holomorphically with the curve; and the Schottky problem identifies the locus of curves inside the locus of all PPAV. The dimension count $g(g+1)/2-(3g-3)=(g-2)(g-3)/2$ is the source of the asymmetry: in low genus the curves fill the candidate space, and in high genus they form a codimension-quadratic subvariety. Schottky's polynomial in genus $4$, the Andreotti-Mayer singular-locus locus, the Novikov-Shiota KP characterisation, and the Welters-Krichever trisecant criterion together constitute the explicit answers; each one connects the geometry of curves to a different mathematical structure (modular forms, theta-divisor singularities, integrable PDEs, projective geometry of Kummer varieties), and the equivalence of these different characterisations is itself a deep theorem of the modern theory. The Krichever construction from algebraic curves to KP solutions, and Shiota's converse extracting the curve from a KP direction, exhibit the Jacobian as a flow-orbit on the Sato Grassmannian, integrating the curve-side geometry with the soliton-side analytic theory.

Full proof set [Master]

Lemma (period-map dimension). For $g\ge 2$, the moduli space ${\mathcal{M}}_g$ of smooth projective complex curves of genus $g$ has complex dimension $3g-3$, and the period map $\mathrm{Per}:{\mathcal{M}}_g\to {\mathcal{A}}_g$ is injective on the locus where the curve is non-hyperelliptic-or-product.

Proof. The dimension count $\dim {\mathcal{M}}_g=3g-3$ for $g\ge 2$ is Riemann's count: the deformation space of a smooth projective curve $X$ of genus $g$ is $H^1(X,T_X)$, of dimension $3g-3$ by Riemann-Roch on the tangent sheaf $T_X$ (degree $2-2g$, so $\dim H^0=0$ for $g\ge 2$ and $\dim H^1=(2g-2)-(1-g)=3g-3$ by Riemann-Roch on $T_X$). Injectivity of $\mathrm{Per}$ is Torelli's theorem (Torelli 1913, modernised by Andreotti 1958 On a theorem of Torelli, Amer. J. Math. 80): a smooth projective curve is determined up to isomorphism by its principally polarised Jacobian. Andreotti's proof for $g\ge 4$ recovers the canonical model of $X$ as the singular locus of the theta divisor; for $g=2,3$ a separate hyperelliptic-aware argument is required. $\square$

Lemma (codimension count). For $g\ge 2$, ${\mathrm{codim}}_{{\mathcal{A}}_g}{\mathcal{J}}_g=(g-2)(g-3)/2$.

Proof. $\dim {\mathcal{A}}_g=\dim {\mathfrak{H}}_g=g(g+1)/2$, and $\dim {\mathcal{J}}_g=\dim {\mathcal{M}}_g=3g-3$ by the previous lemma and the injectivity of the period map. The difference is $g(g+1)/2-(3g-3)=(g^2+g-6g+6)/2=(g^2-5g+6)/2=(g-2)(g-3)/2$. $\square$

Theorem (Schottky's genus-$4$ relation). Statement and proof as in the Intermediate-tier Key theorem section. The Jacobi locus ${\mathcal{J}}_4\subset {\mathcal{A}}_4$ is the irreducible hypersurface cut out by the polynomial $S(\theta )$ of degree $4$ in $12$ specific theta-constants.

Proof. The four-step structure goes through using the canonical curve embedding (a genus-$4$ canonical curve is a complete intersection of a quadric and a cubic in ${\mathbb{P}}^3$), the unique-quadric / two-rulings dichotomy, the resulting two vanishing theta-nulls, and the algebraic relation among the theta-constants from the Riemann-Mumford addition formulae. Igusa 1980 closed the irreducibility step. $\square$

Theorem (Krichever construction, 1977). Given a smooth projective curve $X$ of genus $g$, a marked point $p\in X$ with local coordinate, and a generic line bundle $L$ of degree $g$, the Baker-Akhiezer function $\psi (P,x,y,t):=\theta (\alpha (P)+xV+yU+tW+z_0)/\theta (\alpha (P)+z_0)\cdot \exp (x{f}_x+y{f}_y+t{f}_t)$ on $X\setminus \{p\}$, where $V,U,W\in {\mathbb{C}}^g$ are determined by the local expansion of $\theta$ at $p$ and $f_x,f_y,f_t$ are explicit local functions, is meromorphic on $X\setminus \{p\}$ with prescribed essential singularity at $p$, and the function $u(x,y,t):=-2{\partial }_x^2\log \theta (xV+yU+tW+z_0;\tau )$ solves the KP equation.

Proof. Krichever 1977 Methods of algebraic geometry in the theory of nonlinear equations (Russian Math. Surveys 32) constructs $\psi$ explicitly using the theta function and the local data at $p$. The KP equation for $u$ follows from a residue calculation on $X$, evaluating the higher-derivative trace of the Lax operator $L$ associated to $\psi$. The key computational identity uses the Riemann-Roch formula on $X$ for the line bundles $\mathcal{O}(D+np)$ and the bilinear identity (cf. 06.06.07) on the periods. The proof is approximately three pages of explicit theta-function manipulation; Mulase-Yamada 1995 Solvable structure of nonlinear lattice equations (Comm. Math. Phys. 175) gives the modern Sato-Grassmannian formulation. $\square$

Theorem (Shiota 1986, Novikov conjecture). Statement as in the Advanced results §. The KP-hierarchy condition is necessary and sufficient for $(A,\Theta )$ to be a Jacobian.

Proof. The necessity (Krichever direction) is the previous theorem. Sufficiency (Shiota direction): given a KP-flow direction on $A_{\tau }$, Shiota constructs the spectral curve $X$ of the Lax operator associated to the flow, and proves $A_{\tau }\cong \mathrm{Jac}(X)$ as principally polarised abelian varieties. The construction uses the formal Baker-Akhiezer function (existence forced by the KP equation and the theta function), the residue calculus on the formal disk at the marked point, and the Sato-Grassmannian / KP-tau-function machinery. The proof is approximately fifty pages in Inventiones 83; Arbarello-De Concini 1987 Another proof of a conjecture of S. P. Novikov (Duke Math. J. 54) gives an algebro-geometric proof of approximately twenty pages, using the Brill-Noether geometry of the conjectured curve. $\square$

Theorem (Welters trisecant conjecture, Krichever 2010). A principally polarised abelian variety $(A,\Theta )$ of dimension $g\ge 2$ is a Jacobian iff its Kummer variety $K(\tau )\subset {\mathbb{P}}^{2^g-1}$ admits a trisecant line.

Proof. Necessity (Fay 1973 Theta Functions on Riemann Surfaces, Lecture Notes in Mathematics 352) follows from the Fay trisecant identity for theta-functions on a curve: for any $z_1,z_2,z_3\in A$ with $z_1+z_2+z_3\in \alpha ({\mathrm{Sym}}^3(X))-3\alpha (p_0)$, the second-order theta values are linearly dependent, exhibiting a trisecant of $K(\tau )$. Sufficiency (Krichever 2010, Ann. of Math. 172): a single trisecant produces a one-dimensional family of trisecants by translation in $A$, then a KP-flow direction by tangential limits, then a spectral curve by the Shiota machinery. The argument approximately mirrors Shiota's, with the trisecant condition replacing the KP condition as the input data; Krichever-Phong 2008 Symplectic forms in the theory of solitons gave a unified treatment of trisecant and KP characterisations. $\square$

Theorem (Andreotti-Mayer containment, Andreotti-Mayer 1967). For $g\ge 4$, ${\mathcal{J}}_g\subseteq {\mathcal{N}}_{g-4}$ where ${\mathcal{N}}_k:=\{[A,\Theta ]\in {\mathcal{A}}_g:\dim \mathrm{Sing}(\Theta )\ge k\}$.

Proof. For $X$ a smooth projective curve of genus $g$, the singular locus of the theta divisor $\Theta \subset \mathrm{Jac}(X)$ is identified by Riemann's singularity theorem with the translated Brill-Noether locus $W_{g-1}^1$ of effective line bundles of degree $g-1$ with $h^0\ge 2$. The Brill-Noether expected dimension is $\rho (g,g-1,1)=g-2\cdot (g-(g-1)+1)=g-4$, by the Kempf-Kleiman-Laksov dimension theorem; for a general curve, equality holds by Griffiths-Harris 1980 (cf. 06.06.06), so $\dim \mathrm{Sing}(\Theta )=g-4$ generically and $\dim \mathrm{Sing}(\Theta )\ge g-4$ for every curve. Hence $[\mathrm{Jac}(X),\Theta ]\in {\mathcal{N}}_{g-4}$. The reverse containment is conjectural and false: the theta-null component of ${\mathcal{N}}_{g-4}$ contains products of Jacobians (Beauville 1977); the Prym component contains Pryms of certain double covers (Debarre 1992); and additional components are conjectured but not enumerated. $\square$

Connections [Master]

• Riemann's bilinear relations 06.06.07. The bilinear relations identify the Siegel upper half-space ${\mathfrak{H}}_g$ as the space of candidate period matrices; the Schottky problem starts from this identification and asks which $\tau \in {\mathfrak{H}}_g$ are period matrices of curves. Without (RB1) the matrix would not be symmetric, without (RB2) it would not give a complex torus; with both, it gives a principally polarised abelian variety, and the Schottky problem is the sharpening from "PPAV" to "Jacobian of a curve".

• Variation of Hodge structure on the Jacobian 06.08.02. The variation of Hodge structure is the holomorphic dependence of the period matrix $\tau (X)$ on the moduli parameter $X\in {\mathcal{M}}_g$; the Schottky problem asks for the image of this period mapping. Torelli's theorem makes the map injective; the Schottky problem identifies its image. The two together complete the period-mapping picture: ${\mathcal{M}}_g$ embeds into ${\mathcal{A}}_g$ with image ${\mathcal{J}}_g$, and the Schottky problem characterises ${\mathcal{J}}_g$ explicitly.

• Jacobi inversion theorem 06.06.06. The Jacobi inversion theorem realises every point of the Jacobian as an effective divisor of degree $g$ on the curve, and Riemann's vanishing theorem identifies the theta divisor with the Brill-Noether locus $W_{g-1}$. The Andreotti-Mayer singular-locus characterisation of Jacobians uses Riemann's singularity theorem, identifying $\mathrm{Sing}(\Theta )$ with the translated $W_{g-1}^1$ — the Brill-Noether stratification controls the local geometry of the theta divisor, and the Schottky problem reads this geometry off the abelian variety to recognise Jacobians.

• Theta function 06.06.05. The Riemann theta function $\theta (z;\tau )$ is the holomorphic section of the principal polarisation line bundle on a PPAV; its vanishing locus is the theta divisor. Theta-constants $\theta [m,m^{\prime }](\tau )$ are the values at the half-period characteristics; they are Siegel modular forms of weight $1/2$ on $\Gamma (2)$. Schottky's polynomial $S(\theta )$ is a polynomial relation among theta-constants vanishing exactly on the Jacobi locus; the Shiota-Novikov characterisation uses $\theta (z;\tau )$ as the input to the KP equation.

• Jacobian variety 06.06.03. The Jacobian is the principally polarised abelian variety of a smooth projective curve; the Schottky problem identifies which PPAV are Jacobians. Combined with Torelli's theorem (a curve is determined by its Jacobian), the Schottky locus ${\mathcal{J}}_g\subset {\mathcal{A}}_g$ is the moduli space of curves embedded as a subvariety of the moduli space of PPAV.

• Period matrix 06.06.02. The period matrix is the input data for the Schottky problem: a $g\times 2g$ complex matrix $(A\mid B)$ assembling integrals of holomorphic 1-forms against integral 1-cycles. The bilinear relations cut down to the normalised form $\tau \in {\mathfrak{H}}_g$; the Schottky problem cuts further to the Jacobi locus.

• Holomorphic 1-form 06.06.01. The basis ${\omega }_1,\dots ,{\omega }_g$ of holomorphic 1-forms is the input to the period matrix; the dimension count $\dim H^0(X,{\Omega }_X^1)=g$ is what makes the period matrix $g\times g$. The Schottky problem operates entirely at the level of this period matrix.

• Hodge decomposition on a compact Riemann surface 06.04.03. The Hodge decomposition gives the polarised Hodge structure of weight $1$ on $H^1(X,\mathbb{C})$, which in turn gives the principal polarisation of the Jacobian; the Schottky problem characterises which polarised Hodge structures of this form arise from curves.

• Riemann-Roch theorem for compact Riemann surfaces 06.04.01. Riemann-Roch supplies the Brill-Noether dimension count $\rho (g,d,r)$ that controls the singular locus of the theta divisor and hence the Andreotti-Mayer characterisation of the Jacobi locus.

• Abel-Jacobi map 06.06.04. The Fay trisecant identity, which gives the Jacobian direction of the Welters criterion, follows from the Abel-Jacobi geometry: three points $z_1,z_2,z_3$ on $\mathrm{Jac}(X)$ with $z_1+z_2+z_3$ in the image of ${\mathrm{Sym}}^3(X)$ produce a trisecant of the Kummer variety. The trisecant criterion is the abstract abelian-variety reformulation of this Abel-Jacobi observation.

Historical & philosophical context [Master]

Friedrich Schottky introduced the genus-$4$ relation in the 1888 paper Über die Moduln der Thetafunctionen ^{[Schottky 1888]} (J. Reine Angew. Math. 102, 304-352). Schottky's motivation came from the dimension-count observation: in genus $4$ the moduli space of curves has dimension $9$ while the moduli space of principally polarised abelian fourfolds has dimension $10$, so the Jacobi locus must be a hypersurface. Schottky constructed the polynomial $S(\theta )$ in theta-constants that vanishes on Jacobians, derived from the canonical-curve embedding of a genus-$4$ curve as a complete intersection of a unique quadric and a cubic in ${\mathbb{P}}^3$. The two rulings of the quadric (or one ruling in the cone case) determine the two even theta-characteristics that vanish on the Jacobian's theta divisor; Schottky's relation expresses the algebraic dependence among the resulting theta-constants. Schottky-Jung 1909 Neue Sätze über Symmetralfunktionen und die Abelschen Funktionen extended the framework to higher genus through unramified double covers and the Prym geometry, but the resulting equations are far less explicit than the genus-$4$ case.

The mid-twentieth-century reformulation came from algebraic geometry. Aldo Andreotti 1958 On a theorem of Torelli (Amer. J. Math. 80, 801-828) gave the modern proof of Torelli's theorem (Torelli 1913) using the geometry of the theta divisor: for $g\ge 4$, the singular locus of $\Theta$ recovers the canonical model of the curve, providing both injectivity of the period map and a description of the Jacobian-side data needed for the Schottky problem. Andreotti-Mayer 1967 On period relations for abelian integrals on algebraic curves ^{[Andreotti-Mayer 1967]} (Ann. Sc. Norm. Sup. Pisa 21, 189-238) defined the locus ${\mathcal{N}}_k$ of PPAV whose theta divisor has singular locus of dimension at least $k$, proved ${\mathcal{J}}_g\subseteq {\mathcal{N}}_{g-4}$, and conjectured equality. The conjecture is still open in full generality; Beauville-Debarre 1986 and Debarre 1992 identified additional irreducible components of ${\mathcal{N}}_{g-4}$ in low genus, ruling out naive equality.

The integrable-systems revolution arrived through Sergei Novikov, Igor Krichever, and Takahiro Shiota. Krichever 1977 Methods of algebraic geometry in the theory of nonlinear equations ^{[Krichever 1977]} (Russian Math. Surveys 32, 185-213) constructed explicit solutions of the KP equation from algebraic curves via the Baker-Akhiezer function, extending Mumford-Manin-Lax-Hochstadt-Krichever earlier work on the KdV equation. Sergei Novikov 1979 A periodic problem for the Korteweg-de Vries equation ^{[Novikov 1979]} conjectured that the resulting KP-flow data characterises Jacobians among PPAV. Takahiro Shiota 1986 Characterization of Jacobian varieties in terms of soliton equations ^{[Shiota 1986]} (Invent. Math. 83, 333-382) proved Novikov's conjecture, in approximately fifty pages of Sato-Grassmannian / Lax-pair / Baker-Akhiezer-function calculation. Mikio Sato's 1981 lectures Soliton equations as dynamical systems on infinite-dimensional Grassmann manifold (RIMS Kokyuroku) provided the Grassmannian framework; Mulase 1984 Cohomological structure in soliton equations and Jacobian varieties and Arbarello-De Concini 1987 Another proof of a conjecture of S. P. Novikov ^{[Arbarello-De Concini 1987]} (Duke Math. J. 54, 163-178) gave alternative proofs.

The trisecant criterion came from Gerald Welters and Igor Krichever. Welters 1984 A criterion for Jacobi varieties ^{[Welters 1984]} (Ann. of Math. 120, 497-504) conjectured the criterion and proved it conditional on a regularity assumption; Igor Krichever 2010 Characterizing Jacobians via trisecants of the Kummer variety ^{[Krichever 2010]} (Ann. of Math. 172, 485-516) gave the unconditional proof, twenty-six years after the conjecture. The trisecant identity itself goes back to John Fay 1973 Theta Functions on Riemann Surfaces (Lecture Notes in Mathematics 352), and the Fay identity is one of the foundational identities of the modern theory of Jacobians. David Mumford's Tata Lectures on Theta II ^{[Mumford Tata II]} (Birkhäuser PM 43, 1984) gave the first systematic textbook treatment of the Schottky problem within the modern theta-function theory, including the genus-$4$ Schottky-Jung framework, the Andreotti-Mayer locus, and the Krichever-Novikov KP correspondence. Donaldson's Riemann Surfaces §12-§13 ^{[Donaldson Riemann Surfaces]} presents the Schottky problem in the Hodge-theoretic framework, integrated with the variation of Hodge structure; Arbarello-Cornalba-Griffiths Geometry of Algebraic Curves II (Springer GMW 268, 2011) develops the Brill-Noether-Schottky theory in the modern Hodge-theoretic language.

Bibliography [Master]

[object Promise]