Kodaira embedding theorem

Intuition [Beginner]

The Kodaira embedding theorem answers a foundational question: which compact complex manifolds sit inside projective space? Projective space ${\mathbb{P}}^N$ is the home of algebraic geometry — every projective variety is cut out by polynomial equations in ${\mathbb{P}}^N$. A compact complex manifold that embeds holomorphically into ${\mathbb{P}}^N$ is automatically algebraic. The theorem says: a compact Kähler manifold $X$ admits such an embedding if and only if $X$ carries a positive line bundle.

Kodaira proved this in 1954. The setup is differential-geometric: a line bundle $L$ on $X$ is positive when it admits a Hermitian metric whose curvature is a positive $(1,1)$-form — a closed real $(1,1)$-form that pairs positively with every nonzero tangent direction. Equivalently, the first Chern class $c_1(L)$ is represented by a Kähler form on $X$. From such an $L$, take a very high tensor power $L^k$; its global holomorphic sections become numerous enough to separate every pair of points on $X$ and to distinguish tangent directions. The map sending $x\in X$ to the projective line of sections vanishing at $x$ is then an embedding into a projective space.

The theorem identifies Kähler with a positive line bundle with projective. It is the bridge between complex differential geometry (Kähler manifolds, positivity, curvature) and algebraic geometry (projective varieties, polynomial equations, ample line bundles). Most compact Kähler manifolds are projective, but generic K3 surfaces and certain non-projective complex tori show that the gap is real.

Visual [Beginner]

A compact Kähler manifold with a positive line bundle; the sections of a high power of the line bundle define an embedding into projective space, displayed as a map from the manifold to ${\mathbb{P}}^N$.

The picture shows a compact Kähler surface on the left, the line bundle $L$ sitting above it as fibres, and the embedding into ${\mathbb{P}}^N$ on the right. The dimensions of the fibres of $L^k$ over $X$ grow polynomially in $k$, supplying the sections needed to separate points and tangent vectors.

Worked example [Beginner]

For $X=$ a smooth projective curve of genus 2 (compact Riemann surface), the canonical line bundle ${\omega }_X$ has degree $2g-2=2$. The canonical bundle is positive: it admits a metric of positive curvature once $g\ge 2$. (For $g=1$ the canonical bundle is the structure sheaf, which is not positive but is flat; for $g=0$ the canonical is negative.)

Step 1. The space of global holomorphic sections of the canonical ${\omega }_X$ has dimension $g=2$ — there are 2 independent holomorphic 1-forms.

Step 2. The canonical itself does not yet embed $X$ into ${\mathbb{P}}^1$, since the canonical map $X\to {\mathbb{P}}^1$ from a genus-2 curve is the hyperelliptic involution: a degree-2 cover of ${\mathbb{P}}^1$, not an embedding.

Step 3. The 3rd tensor power of the canonical bundle, written ${\omega }_X^3$, has degree $6$ and by Riemann-Roch its space of global sections has dimension $\mathrm{deg}+1-g=6+1-2=5$. The associated map $X\to {\mathbb{P}}^4$ sends each point of $X$ to a projective line in a 5-dimensional space of sections.

Step 4. Verify embedding: the map separates points (no two points map to the same line of sections, since sections cut out divisors of degree 6 and any point can be distinguished from any other by some section vanishing at one and not the other) and separates tangents (sections vanishing at $x$ to second order are fewer than sections vanishing only at $x$).

What this tells us: a positive line bundle and its high tensor powers eventually give an honest embedding; for the genus-2 curve, the 3rd canonical power ${\omega }_X^3$ already embeds.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $X$ be a compact complex manifold of complex dimension $n$. A Hermitian line bundle on $X$ is a holomorphic line bundle $L\to X$ together with a Hermitian metric $h$ — a smoothly varying positive-definite Hermitian inner product on the fibres of $L$.

Definition (positive line bundle). A Hermitian line bundle $(L,h)$ on a complex manifold $X$ is positive if its Chern curvature

$$i\Theta (L,h)=-i\partial \bar{\partial }\log \Vert s{\Vert }_h^2$$

(computed using any local non-vanishing holomorphic section $s$ of $L$) is a positive $(1,1)$-form on $X$ — i.e., $i\Theta (L,h)$ is a real $(1,1)$-form that is positive-definite as a Hermitian form on the holomorphic tangent space $T_X^{1,0}$ at every point.

A line bundle $L$ is positive if it admits a positive Hermitian metric.

Definition (Hodge class, Kähler class). Let $X$ be a compact Kähler manifold of dimension $n$ with a fixed Kähler form $\omega$. A class $\alpha \in H^2(X;\mathbb{Z})$ whose image in $H^2(X;\mathbb{C})={⨁}_{p+q=2}{H}^{p,q}$ lies entirely in $H^{1,1}(X)$ is called a Hodge class of type $(1,1)$. A class $\alpha \in H^{1,1}(X;\mathbb{R}):=H^{1,1}(X)\cap H^2(X;\mathbb{R})$ that has a positive (Kähler) representative is called a Kähler class.

Theorem (Kodaira embedding theorem). Let $X$ be a compact complex manifold. The following are equivalent:

(K1) $X$ is projective: there exists a holomorphic embedding $X\hookrightarrow {\mathbb{P}}^N$ for some $N$.

(K2) $X$ is Kähler and admits a positive holomorphic line bundle $L$.

(K3) $X$ is Kähler and admits an integral Kähler class — i.e., a Kähler class $[\omega ]\in H^{1,1}(X;\mathbb{R})$ that lies in the image of $H^2(X;\mathbb{Z})\to H^2(X;\mathbb{R})$.

The line-bundle and Hodge-class formulations are equivalent. The bridge is the Lefschetz $(1,1)$-theorem: every Hodge class of type $(1,1)$ on a compact Kähler manifold is the first Chern class $c_1(L)$ of some holomorphic line bundle $L$. The positive case picks out exactly the Kähler classes. The Lefschetz theorem appears in 04.09.09 in the catalog; its proof uses the exponential exact sequence

$$0\to \mathbb{Z}\to {\mathcal{O}}_X\stackrel{\exp (2\pi i\cdot )}{\to }{\mathcal{O}}_X^{\times }\to 0$$

and the resulting long exact sequence, identifying $H^1(X;{\mathcal{O}}_X^{\times })=\mathrm{Pic}(X)$ with the kernel of $H^2(X;\mathbb{Z})\to H^2(X;{\mathcal{O}}_X)$.

Counterexamples to common slips [Intermediate+]

• Kähler does not imply projective. A generic K3 surface (in the analytic moduli space of K3s) is Kähler but not projective. The obstruction is integrality: the Kähler cone exists inside $H^{1,1}(X;\mathbb{R})$, but $H^{1,1}\cap H^2(X;\mathbb{Z})$ may be reduced to zero in the generic complex structure. The Noether-Lefschetz locus inside the moduli of K3s is the set of complex structures where Hodge classes appear, and it is everywhere dense but of measure zero.
• Not all compact complex manifolds are Kähler. Hopf surfaces $({\mathbb{C}}^2\setminus \{0\})/⟨z\mapsto \lambda z⟩$ for $|\lambda |\ne 1$ are compact complex surfaces diffeomorphic to $S^1\times S^3$. Their first Betti number is $1$, an odd number, ruling out a Kähler metric (Kähler forces $b_1$ to be even, since $b_1=2h^{1,0}$). Hence Hopf surfaces are neither Kähler nor projective.
• Moishezon manifolds need not be Kähler. A Moishezon manifold is a compact complex manifold $X$ with $n$ algebraically independent meromorphic functions (so the field of meromorphic functions has transcendence degree $n={\dim }_{\mathbb{C}}X$). Every projective variety is Moishezon. The converse fails: there exist Moishezon non-Kähler threefolds. Chow's theorem (1949) says every closed analytic subspace of ${\mathbb{P}}^N$ is algebraic; Moishezon's theorem (1966) characterises the algebraic-equivalent class of compact complex manifolds in terms of bimeromorphic equivalence to a projective variety.

Key theorem with proof [Intermediate+]

Theorem (Kodaira embedding theorem). Let $X$ be a compact Kähler manifold of complex dimension $n$ admitting a positive holomorphic line bundle $L$. Then there exists $k_0=k_0(X,L)$ such that for all $k\ge k_0$, the linear system $|L^{\otimes k}|$ defines a holomorphic embedding

$${\phi }_{L^k}:X\hookrightarrow \mathbb{P}(H^0(X;L^{\otimes k}{)}^{\vee })={\mathbb{P}}^{N_k},N_k=h^0(X;L^{\otimes k})-1.$$

In particular, $X$ is a smooth projective variety.

Proof. The proof proceeds in five steps. The architecture is: Kodaira vanishing $\Rightarrow$ sufficiently many sections $\Rightarrow$ separation of points $\Rightarrow$ separation of tangent vectors $\Rightarrow$ embedding.

Step 1 — Setup of the map ${\phi }_{L^k}$. For a line bundle $L^{\otimes k}$ with global sections $s_0,\dots ,s_{N_k}$ forming a basis of $H^0(X;L^{\otimes k})$, define the Kodaira map

$${\phi }_{L^k}:X\setminus \mathrm{Bs}(L^k)\to {\mathbb{P}}^{N_k},x\mapsto [s_0(x):s_1(x):\cdots :s_{N_k}(x)],$$

where $\mathrm{Bs}(L^k)=\{x\in X:s(x)=0\text{ for all }s\in H^0(L^{\otimes k})\}$ is the base locus. To prove the embedding, we must show: (a) for $k$ large, $\mathrm{Bs}(L^k)=\varnothing$; (b) ${\phi }_{L^k}$ is injective; (c) $d{\phi }_{L^k}$ is everywhere injective.

Step 2 — Vanishing input from Kodaira. For a positive line bundle $L$ on $X$, the line bundle $L^{\otimes k}\otimes {\omega }_X^{-1}$ is also positive for $k$ sufficiently large (since ${\omega }_X^{-1}$ is a fixed bundle and $L^{\otimes k}$ has curvature growing linearly in $k$). By the Kodaira vanishing theorem 04.09.02 applied to $L^{\otimes k}\otimes {\omega }_X^{-1}$:

$$H^i(X;{\omega }_X\otimes L^{\otimes k}\otimes {\omega }_X^{-1})=H^i(X;L^{\otimes k})=0,i>0.$$

So $h^i(L^{\otimes k})=0$ for $i>0$ and $k$ sufficiently large. By the Hirzebruch-Riemann-Roch theorem, $h^0(X;L^{\otimes k})=\chi (X;L^{\otimes k})$ grows like $(L^n/n!)k^n$ — a polynomial of degree $n$ with positive leading coefficient $L^n/n!>0$ (since $L$ is positive, the self-intersection $L^n$ is positive).

Step 3 — Separation of points and base-point-freeness. Let $p,q\in X$ be distinct points and let ${\mathcal{I}}_{p,q}$ denote the ideal sheaf vanishing at both. We have the short exact sequence

$$0\to L^{\otimes k}\otimes {\mathcal{I}}_{p,q}\to L^{\otimes k}\to L^{\otimes k}{|}_{\{p,q\}}\to 0.$$

Taking global sections gives the long exact sequence:

$$0\to H^0(L^{\otimes k}\otimes {\mathcal{I}}_{p,q})\to H^0(L^{\otimes k})\to L_p^{\otimes k}\oplus L_q^{\otimes k}\to H^1(L^{\otimes k}\otimes {\mathcal{I}}_{p,q})\to \cdots$$

By Kodaira vanishing applied to $L^{\otimes k}\otimes {\omega }_X^{-1}\otimes {\mathcal{I}}_{p,q}$ (with appropriate $k$-shift so that $L^{\otimes k}\otimes {\omega }_X^{-1}$ remains positive after the ideal twist; this requires a multiplier ideal refinement, or alternatively a Castelnuovo-Mumford regularity bound): for $k$ large,

$$H^1(X;L^{\otimes k}\otimes {\mathcal{I}}_{p,q})=0.$$

Hence the restriction map $H^0(L^{\otimes k})\to L_p^{\otimes k}\oplus L_q^{\otimes k}$ is surjective. Choosing a section vanishing at $p$ but not at $q$ separates the points: ${\phi }_{L^k}(p)\ne {\phi }_{L^k}(q)$. The case $p=q$ also gives $\mathrm{Bs}(L^k)=\varnothing$ (sections do not all vanish at any single point).

Step 4 — Separation of tangent vectors. For a tangent vector $v\in T_{p}X$, $v\ne 0$, we want to show that some section vanishes at $p$ with derivative $\ne 0$ in direction $v$. Use the exact sequence

$$0\to L^{\otimes k}\otimes {\mathfrak{m}}_p^2\to L^{\otimes k}\otimes {\mathfrak{m}}_p\to T_p^{\vee }X\otimes L_p^{\otimes k}\to 0$$

(after twisting). By Kodaira vanishing applied to $L^{\otimes k}\otimes {\omega }_X^{-1}\otimes {\mathfrak{m}}_p^2$, again for $k$ large,

$$H^1(X;L^{\otimes k}\otimes {\mathfrak{m}}_p^2)=0,$$

and the restriction $H^0(L^{\otimes k}\otimes {\mathfrak{m}}_p)\to T_p^{\vee }X\otimes L_p^{\otimes k}$ is surjective. So given any tangent direction $v$, some section has first-order vanishing at $p$ with non-zero derivative in $v$. Hence $d{\phi }_{L^k}{|}_p$ is injective on $T_{p}X$.

Step 5 — Conclusion. Combining Steps 1–4: for $k\ge k_0$ (uniform over $X$ by compactness), ${\phi }_{L^k}$ is base-point-free, injective on points, and injective on tangent vectors. So ${\phi }_{L^k}:X\to {\mathbb{P}}^{N_k}$ is a holomorphic immersion that is bijective onto its image. Since $X$ is compact and the image is a closed analytic subset of ${\mathbb{P}}^{N_k}$, the map is a homeomorphism onto a closed subset. The image is a closed complex submanifold, hence (by Chow's theorem, 1949) a smooth algebraic subvariety of ${\mathbb{P}}^{N_k}$. $\square$

Bridge. The construction here builds toward 04.09.10 — Akizuki-Nakano vanishing strengthens Kodaira and tightens the bound $k_0$ via multiplier-ideal techniques. The same separation argument appears again in the proof of very ampleness criteria in birational geometry. Putting these together, the foundational reason that positivity yields embedding is the curvature-vanishing pairing: positive curvature of $L$ forces $\bar{\partial }$-harmonic forms of higher degree on ${\omega }_X\otimes L^k$ to be zero, which is exactly what makes the restriction maps in the long exact sequences surjective. The bridge is between analytic positivity (Hermitian curvature) and algebraic embedding (sections of high tensor powers separating geometric data).

Exercises [Intermediate+]

Lean formalization [Intermediate+]

lean_status: partial — Mathlib hosts projective schemes, ample line bundles, the Picard group of a scheme, and partial sheaf-cohomology infrastructure. The Kähler-manifold side (positive Hermitian metrics on holomorphic line bundles, harmonic-form representatives, the Chern curvature as a $(1,1)$-form) is not yet present at the depth required for a proof. The Kodaira vanishing theorem itself is partial in 04.09.02, so the present module states the embedding theorem and the separation-of-points / separation-of-tangents lemmas with sorry proofs to give downstream files an anchor.

import Mathlib.AlgebraicGeometry.Scheme

namespace Codex.AlgGeom.Hodge

-- Kodaira embedding (Kodaira 1954): a compact Kähler manifold X is
-- projective iff X admits a positive holomorphic line bundle L.
-- Equivalently iff X carries a Hodge class of type (1,1) (an integral
-- (1,1)-class with positive (1,1)-form representative).
--
-- The proof factors through Kodaira vanishing: high tensor powers L^k
-- have enough global sections (Kodaira vanishing kills H^1 of twists
-- by ideal sheaves I_p, I_p^2) to separate points and tangent vectors,
-- producing the embedding into P^N where N = h^0(X, L^k) - 1.

-- See lean/Codex/AlgGeom/Hodge/KodairaEmbedding.lean for theorem
-- statements: kodaira_embedding, kodaira_embedding_hodge,
-- high_power_separates_points, high_power_separates_tangents.

end Codex.AlgGeom.Hodge

The module declares predicates KahlerManifold, LineBundle, IsPositive, IsProjective, and HodgeClass as placeholders pending the Mathlib infrastructure described above. The named theorem statements kodaira_embedding, kodaira_embedding_hodge, high_power_separates_points, and high_power_separates_tangents are recorded with sorry proof bodies; their statements track the mathematical content of the embedding theorem and its technical inputs.

The lean_mathlib_gap for this module is non-empty: missing are the Kähler-manifold structure (Mathlib's Geometry.Manifold lacks Hermitian holomorphic vector bundles with curvature), the positive $(1,1)$-form / Chern-curvature framework, Hodge classes as the kernel of the exponential connecting homomorphism, and the full proof of Kodaira vanishing on which the embedding-theorem proof depends. Closing these gaps is the substantive Lean contribution and would lift lean_status from partial to full.

Advanced results [Master]

Proof via Kodaira vanishing — high powers of $L$ separate points and tangents [Master]

The core of the Kodaira embedding theorem is the separation-of-points and separation-of-tangents statements proved via Kodaira vanishing. The argument deserves a self-contained synthesis.

Lemma (separation of points via Kodaira). Let $X$ be a compact Kähler manifold of complex dimension $n$ with a positive line bundle $L$. For all $k$ sufficiently large and any two distinct points $p,q\in X$, the restriction map

$$H^0(X;L^{\otimes k})\to L_p^{\otimes k}\oplus L_q^{\otimes k}$$

is surjective. Consequently, sections of $L^{\otimes k}$ separate $p$ and $q$.

Proof. Consider the ideal sheaf ${\mathcal{I}}_{p,q}$ of the reduced subscheme $\{p,q\}$ inside $X$. The short exact sequence

$$0\to L^{\otimes k}\otimes {\mathcal{I}}_{p,q}\to L^{\otimes k}\to L^{\otimes k}{|}_{\{p,q\}}\to 0$$

gives in cohomology

$$H^0(L^{\otimes k})\to L_p^{\otimes k}\oplus L_q^{\otimes k}\to H^1(L^{\otimes k}\otimes {\mathcal{I}}_{p,q})\to H^1(L^{\otimes k}).$$

For $k$ large, $L^{\otimes k}\otimes {\omega }_X^{-1}$ is positive (sum of a positive line bundle with positive multiplier and a fixed bundle, dominating asymptotically), so by Kodaira vanishing $H^1(L^{\otimes k})=0$. Treating the twist by ${\mathcal{I}}_{p,q}$ requires either multiplier ideal sheaf refinement (Nadel vanishing 1990) or a Castelnuovo-Mumford regularity bound: for $k\ge k_0(p,q)$ depending on a chosen Hermitian metric and the local geometry near $p,q$, we have $H^1(L^{\otimes k}\otimes {\mathcal{I}}_{p,q})=0$. By compactness of $X\times X\setminus {\Delta }_X$, a single uniform $k_0$ works for all pairs $(p,q)$. Hence the restriction map is surjective. $\square$

Lemma (separation of tangents via Kodaira). Under the same hypotheses, for $k$ sufficiently large and any point $p\in X$ with non-zero tangent vector $v\in T_{p}X$, there is a section of $L^{\otimes k}$ vanishing at $p$ with non-zero derivative in direction $v$.

Proof. Use the second-order ideal sheaf ${\mathfrak{m}}_p^2$. The sequence

$$0\to L^{\otimes k}\otimes {\mathfrak{m}}_p^2\to L^{\otimes k}\otimes {\mathfrak{m}}_p\to L_p^{\otimes k}\otimes T_p^{\vee }X\to 0$$

gives in cohomology

$$H^0(L^{\otimes k}\otimes {\mathfrak{m}}_p)\to L_p^{\otimes k}\otimes T_p^{\vee }X\to H^1(L^{\otimes k}\otimes {\mathfrak{m}}_p^2).$$

By Kodaira vanishing applied to $L^{\otimes k}\otimes {\omega }_X^{-1}\otimes {\mathfrak{m}}_p^2$ (positive for $k$ large by the same asymptotic argument), $H^1(L^{\otimes k}\otimes {\mathfrak{m}}_p^2)=0$. The restriction is surjective onto $L_p^{\otimes k}\otimes T_p^{\vee }X$. Given any $v\in T_{p}X$ and a non-zero element of $L_p^{\otimes k}$, the surjectivity yields a section with the required first-order behaviour. $\square$

Combining the two lemmas with the immersion criterion (a holomorphic map $\phi :X\to {\mathbb{P}}^N$ is an embedding iff it is injective and $d\phi$ is everywhere injective) yields the Kodaira embedding theorem. The architecture is uniform: Kodaira vanishing kills exactly the obstruction to surjectivity of restriction maps on cohomology.

The Hodge class as the obstruction — Kähler + Hodge class implies projective [Master]

The embedding theorem refactors through the integral Kähler-class statement (K3) of the main theorem. The bridge between line bundles and integral cohomology classes is the Lefschetz $(1,1)$-theorem [04.09.09 sibling pending], which says: every integral cohomology class of type $(1,1)$ on a compact Kähler manifold is the first Chern class of a holomorphic line bundle.

Hodge class. A class $\alpha \in H^2(X;\mathbb{Z})$ is a Hodge class of type $(1,1)$ when its image in $H^2(X;\mathbb{C})$ lies entirely in $H^{1,1}(X)$. By the Lefschetz $(1,1)$-theorem, the Hodge classes of type $(1,1)$ are exactly $c_1(\mathrm{Pic}(X))\subset H^2(X;\mathbb{Z})$.

Positivity of a Hodge class. A Hodge class is positive if some real representative in $H^{1,1}(X;\mathbb{R})$ is a positive $(1,1)$-form — i.e., a Kähler form. The space of positive real $(1,1)$-classes is an open cone (the Kähler cone $K(X)\subset H^{1,1}(X;\mathbb{R})$), and its intersection with the integral lattice $H^2(X;\mathbb{Z})$ inside $H^{1,1}(X;\mathbb{R})$ is the set of positive Hodge classes.

Reformulation of Kodaira embedding. $X$ is projective iff $X$ carries a positive Hodge class iff the Kähler cone $K(X)\cap H^2(X;\mathbb{Z})$ is non-empty inside $H^{1,1}(X;\mathbb{R})$.

Why integrality is the obstruction. The Kähler cone is always non-empty for a Kähler manifold (the Kähler form itself is in the cone). The integral lattice $H^2(X;\mathbb{Z})$ inside $H^2(X;\mathbb{R})$ is fixed by the underlying topology. As the complex structure on $X$ varies, the subspace $H^{1,1}(X;\mathbb{R})\subset H^2(X;\mathbb{R})$ varies — and the question is whether the variable subspace passes through any integral points within the Kähler cone.

For a K3 surface: ${\dim }_{\mathbb{R}}{H}^2=22$, ${\dim }_{\mathbb{R}}{H}^{1,1}=20$. The Kähler cone has real dimension 20. As complex structure varies (over a 20-dimensional moduli), the 20-dimensional subspace $H^{1,1}$ sweeps out the Grassmannian. Generic positioning misses the rational sub-lattice, so generic K3 has Picard rank 0. The Noether-Lefschetz locus is the countable union of codimension-1 sublocii where $H^{1,1}$ contains rational classes; on the union, the K3 is projective.

Density of the Noether-Lefschetz locus. The Noether-Lefschetz locus is a countable union of analytic hypersurfaces (closed and proper); its complement is dense in the moduli (a Baire generic K3 is non-projective). But the locus itself is also dense (every K3 has projective deformations arbitrarily close in moduli). The picture is exactly the one for Diophantine approximation: rational points are dense and yet measure-zero.

Application — Calabi-Yau case. For a smooth projective Calabi-Yau $n$-fold $X$ with ${\omega }_X={\mathcal{O}}_X$, the Kähler cone in $H^{1,1}(X;\mathbb{R})$ is open. The integrality condition is substantive: Calabi-Yau 3-folds can have $h^{1,1}$ as large as several hundred, and projective Calabi-Yaus pick out a sub-lattice inside.

Worked example — general K3 surface — Kähler not projective for generic complex structure [Master]

The K3 surface is the textbook obstruction to Kodaira embedding turning into "Kähler $\Rightarrow$ projective" automatically.

Definition (K3 surface). A K3 surface is a compact complex surface $X$ satisfying:

• ${\omega }_X={\mathcal{O}}_X$ (canonical bundle is the structure sheaf).
• $h^{1,0}(X)=0$ (no holomorphic 1-forms).

These conditions force the Hodge diamond

$$\begin{array}{ccccc}& & 1& & \\ & 0& & 0& \\ 1& & 20& & 1\\ & 0& & 0& \\ & & 1& & \end{array}$$

with Betti numbers $b_0=b_4=1$, $b_1=b_3=0$, $b_2=22$. The integral cohomology $H^2(X;\mathbb{Z})$ is the unimodular even lattice of rank 22 and signature $(3,19)$.

K3 surfaces exist and are Kähler. Examples include smooth quartic hypersurfaces $X\subset {\mathbb{P}}^3$ ($X=\{F=0\}$ for a generic homogeneous quartic polynomial $F$), Kummer surfaces (quotients of complex tori by $\pm 1$), and elliptic fibrations of specific types. All algebraic K3 examples are projective; the analytic K3 surfaces include non-projective ones.

Period domain and the local Torelli theorem. The period domain of K3 surfaces is

$$\mathcal{D}=\{[\omega ]\in \mathbb{P}(L_{\mathbb{C}}):\omega \cdot \omega =0,\omega \cdot \bar{\omega }>0\}$$

where $L=U^{\oplus 3}\oplus E_8(-1{)}^{\oplus 2}$ is the K3 lattice. This $\mathcal{D}$ has complex dimension 20 (it parametrises the choice of $H^{2,0}\subset H^2(X;\mathbb{C})$ subject to the Hodge-Riemann bilinear relations).

The local Torelli theorem (Piateski-Shapiro-Šafarevič 1971; Burns-Rapoport 1975). The period map (from the moduli space of K3 surfaces to $\mathcal{D}$) is locally a biholomorphism — the K3 surface is determined up to isomorphism by its Hodge structure on $H^2$.

Generic K3 fails to be projective. The Picard rank $\rho (X)$ equals the rank of the lattice $\mathrm{Pic}(X)=H^{1,1}(X)\cap H^2(X;\mathbb{Z})$. For a generic $\omega \in \mathcal{D}$ (meaning generic complex structure on K3), the subspace $H^{2,0}(X)=\mathbb{C}\cdot \omega$ is orthogonal (in the lattice sense) to no integral vectors. The orthogonal complement in $L_{\mathbb{R}}$ generically contains no integral vectors except 0. Hence $\rho =0$ and the K3 carries no non-zero holomorphic line bundle, hence no positive line bundle, hence no projective embedding.

Jump to projective on the Noether-Lefschetz locus. A K3 acquires an extra Hodge class precisely when the chosen complex structure $\omega$ becomes orthogonal to a non-zero integral vector $\alpha \in L$. The condition $\omega \cdot \alpha =0$ defines a complex hyperplane in $L_{\mathbb{C}}$, intersecting the period domain $\mathcal{D}$ in a complex hypersurface — the Noether-Lefschetz divisor ${\mathcal{D}}_{\alpha }\subset \mathcal{D}$.

The union ${\bigcup }_{\alpha }{\mathcal{D}}_{\alpha }$ (over non-zero integral classes $\alpha$) is the Noether-Lefschetz locus — a countable union of complex hypersurfaces inside $\mathcal{D}$, of measure zero but dense (every $\omega$ is the limit of a sequence in ${\bigcup }_{\alpha }{\mathcal{D}}_{\alpha }$).

On the Noether-Lefschetz locus the K3 has $\rho \ge 1$ and admits at least one holomorphic line bundle. If the line bundle has positive self-intersection (the Hodge index is forced to be either positive or negative on the orthogonal lattice), the K3 is projective. By a standard argument, $\rho \ge 1$ on a K3 with positive Hodge class gives projectivity.

Numerical signature. A K3 with $\rho =1$ is minimally projective — it carries a single line bundle generator. As $\rho$ increases (up to a maximum of 20, the singular K3 case), the geometry becomes more rigid. Algebraic K3s of various $\rho$ have been studied extensively, with the Kuga-Satake construction embedding K3s into moduli of abelian varieties through their Hodge structures.

Picard-Fuchs interpretation. Variation of Hodge structure across the moduli space of K3 surfaces is governed by the Picard-Fuchs equations — a system of linear ODEs in the complex moduli parameter, encoding how the integral cohomology classes split into Hodge components as the complex structure varies. The Noether-Lefschetz loci are the singular sets of these Picard-Fuchs equations.

Moduli-theoretic consequence — moduli of polarised varieties as algebraic spaces [Master]

The Kodaira embedding theorem unlocks the moduli theory of polarised varieties as a fundamentally algebraic-geometric object, distinct from the Kähler-only moduli of complex structures.

Polarised variety. A polarised compact complex manifold $(X,[\omega ])$ is a compact complex manifold $X$ together with a chosen Hodge class $[\omega ]\in H^{1,1}(X;\mathbb{R})\cap H^2(X;\mathbb{Z})$ inside the Kähler cone. By Kodaira embedding, a polarised compact Kähler manifold is automatically projective, and the polarisation $[\omega ]$ is the cohomology class of a hyperplane section under the embedding.

Moduli of polarised varieties. Fix discrete data: dimension $n$, Hodge polynomial $h^{p,q}$, and a polarisation type (intersection numbers of $[\omega {]}^k$). The moduli space ${\mathcal{M}}_{n,h,[\omega ]}$ of polarised varieties with this discrete data is a complex analytic space (parametrising isomorphism classes of polarised pairs $(X,[\omega ])$). For compact families, $\mathcal{M}$ is a Hausdorff complex space.

Algebraicity by Kodaira embedding. Because a polarised pair $(X,[\omega ])$ is projective (Kodaira), the moduli space ${\mathcal{M}}_{n,h,[\omega ]}$ is algebraic: it has the structure of an algebraic variety (or algebraic stack, for objects with non-identity automorphisms). This is the core consequence of Kodaira embedding for moduli theory.

The construction: any family $\pi :\mathcal{X}\to S$ of polarised varieties yields, by Kodaira embedding applied fibrewise, a family of projective embeddings $\mathcal{X}\to S\times {\mathbb{P}}^N$, whose image is an algebraic family. The moduli of such families is then a quotient of a Hilbert scheme, an explicitly constructible algebraic stack.

Classical examples.

• Moduli of polarised curves. For $X$ a smooth projective curve of genus $g$ and $[\omega ]=$ the canonical (assuming $g\ge 2$), the polarised pair is determined by the curve. The moduli space ${\mathcal{M}}_g$ is an algebraic stack of dimension $3g-3$ over $\mathbb{C}$.
• Moduli of polarised K3 surfaces. The moduli of K3 surfaces with a fixed polarisation of degree $2d$ (a class $[\omega ]\in H^2(X;\mathbb{Z})$ with $[\omega {]}^2=2d$, primitive) is a 19-dimensional quasi-projective variety ${\mathcal{M}}_{2d}$ — a Shimura variety of orthogonal type. The full moduli of all K3s (without polarisation) is not algebraic in this sense; it is a 20-dimensional analytic space whose algebraic structure depends on the polarisation choice.
• Moduli of polarised abelian varieties. The moduli space ${\mathcal{A}}_g$ of principally polarised abelian varieties of dimension $g$ is an $\left(\genfrac{}{}{0ex}{}{g+1}{2}\right)$-dimensional Shimura variety, the Siegel modular variety.
• Moduli of polarised Calabi-Yau manifolds. For Calabi-Yau $n$-folds with ${\omega }_X={\mathcal{O}}_X$, the moduli of polarised pairs is a $h^{n-1,1}(X)$-dimensional variety — this dimension is the complex moduli of Calabi-Yau, classically $h^{1,1}$ for surfaces (K3: 19) and $h^{2,1}$ for threefolds (quintic Calabi-Yau: 101).

Distinguished feature. The polarised moduli are algebraic; the unpolarised moduli of complex structures are only complex analytic. This dichotomy — polarised algebraic vs unpolarised complex-analytic — is the foundational signature of Kodaira embedding.

Hodge-theoretic refinement. Each polarised moduli space carries a variation of Hodge structure: the cohomology $H^k(X_s;\mathbb{Q})$ of the fibre varies as a local system on the moduli space, equipped with the Hodge filtration. The period map $\mathcal{M}\to \mathcal{D}/\Gamma$ (to the period domain modulo arithmetic group action) encodes this variation. The Torelli theorems (local and global) ask whether the period map is injective — true for K3s, true for curves of genus $\ge 2$, often subtle in general.

Polarised birational geometry. The minimal model programme of Mori (1982), pursued through Kawamata-Viehweg vanishing (a Kodaira-vanishing extension), produces minimal models of polarised varieties — birational equivalents with mild positivity conditions on the canonical. The output is again polarised; Kodaira embedding ensures the resulting minimal models are projective.

Synthesis. The Kodaira embedding theorem identifies compact Kähler manifolds carrying a positive line bundle (or equivalently a positive Hodge class) with smooth projective varieties, and this identification is the foundational reason that complex algebraic geometry sits inside complex differential geometry as a positivity condition rather than a separate phenomenon. The central insight is that positivity of curvature (an analytic notion) translates, through Kodaira vanishing of higher cohomology, into plenty of sections (an algebraic notion); putting these together via the separation-of-points and separation-of-tangents arguments gives the embedding. This is exactly the bridge between analysis and algebra that makes complex projective geometry a single subject. The Lefschetz $(1,1)$-theorem identifies integral Hodge classes of type $(1,1)$ with holomorphic line bundles, and the Kodaira cone of positive Hodge classes generalises to the ample cone in $\mathrm{Pic}(X)\otimes \mathbb{R}$, dual to the Mori cone of effective curves. The pattern recurs across the modern landscape: Moishezon's theorem identifies Kähler-Moishezon manifolds with projective ones; Yau's solution of the Calabi conjecture (1978) identifies positive Hodge classes on Calabi-Yau manifolds with Ricci-flat Kähler metrics; the Kuga-Satake construction embeds K3 surfaces into moduli of abelian varieties via their polarised Hodge structures; and p-adic Hodge theory (Fontaine-Faltings) carries the polarised-Hodge picture into arithmetic, where the period domain $\mathcal{D}$ becomes a p-adic period domain. Across all these reframings, Kodaira embedding is the foundational positivity criterion: positive line bundle $\iff$ projective embedding $\iff$ polarised algebraic moduli.

Full proof set [Master]

Proposition 1 (Kodaira embedding theorem). Let $X$ be a compact Kähler manifold admitting a positive holomorphic line bundle $L$. Then for $k$ sufficiently large, the Kodaira map ${\phi }_{L^k}:X\to \mathbb{P}(H^0(X;L^{\otimes k}{)}^{\vee })$ is a holomorphic embedding, and $X$ is a smooth projective variety.

Proof. This is proved in the Key Theorem section in five steps. The architecture: Kodaira vanishing for $L^{\otimes k}\otimes {\omega }_X^{-1}$ for $k$ large yields surjectivity of restriction maps from $H^0(L^{\otimes k})$ to fibrewise and infinitesimal data at points of $X$. Combining surjectivity at distinct points (separation of points), with surjectivity at second-order infinitesimal neighbourhoods of single points (separation of tangents), and the compactness of $X$, gives that ${\phi }_{L^k}$ is a holomorphic embedding. Chow's theorem then converts the embedded image into an algebraic subvariety of ${\mathbb{P}}^{N_k}$. $\square$

Proposition 2 (Positivity-projectivity equivalence). A compact Kähler manifold $X$ is projective iff $X$ admits a positive holomorphic line bundle iff $X$ carries an integral Kähler class.

Proof. The forward direction "projective $\Rightarrow$ positive line bundle and Kähler" is Exercise 4: pull back the Fubini-Study form. The reverse direction "positive line bundle $\Rightarrow$ projective" is Proposition 1. The equivalence between positive line bundle and integral Kähler class is the Lefschetz $(1,1)$-theorem (Exercise 6): an integral Hodge class of type $(1,1)$ is the first Chern class of a holomorphic line bundle, and positivity of the line bundle corresponds to a positive representative of the cohomology class — i.e., a Kähler class. $\square$

Proposition 3 (Sharpness — Kähler does not imply projective). There exist compact Kähler manifolds that are not projective. The generic K3 surface (with generic complex structure in the analytic moduli) is such an example.

Proof. For a K3 surface $X$ with generic complex structure $\omega \in \mathcal{D}$ (in the 20-dimensional period domain), the subspace $H^{1,1}(X;\mathbb{R})={\omega }^{\perp }\cap L_{\mathbb{R}}$ has real dimension 20 inside $L_{\mathbb{R}}\cong {\mathbb{R}}^{22}$. As $\omega$ varies in the period domain $\mathcal{D}$, the subspace $H^{1,1}$ sweeps out a 20-dimensional submanifold of the Grassmannian $\mathrm{Gr}(20,22)$. Generic positioning of this submanifold misses every rational subspace: the rational subspaces in ${\mathbb{R}}^{22}$ are a countable union of $\mathbb{Q}$-vector subspaces, of measure zero. So generically the intersection $H^{1,1}(X;\mathbb{R})\cap H^2(X;\mathbb{Q})$ is reduced to $\{0\}$.

Equivalently: $\mathrm{Pic}(X)=H^{1,1}(X)\cap H^2(X;\mathbb{Z})=0$ for generic $\omega$. No non-zero holomorphic line bundle exists on the generic K3, so no positive line bundle, and by Kodaira embedding (or Lefschetz $(1,1)$), the generic K3 is not projective. $\square$

Proposition 4 (Moduli-theoretic consequence). Let $(X,[\omega ])$ be a polarised compact Kähler manifold with $[\omega ]$ a positive Hodge class. Then the moduli space ${\mathcal{M}}_{n,h,[\omega ]}$ of polarised pairs with the discrete data $(n=\dim X,h=$ Hodge polynomial, $[\omega ]=$ polarisation type$)$ is an algebraic space (Artin algebraic stack).

Proof. By Proposition 2, every polarised pair is projective. Fix a sufficiently large $k$ so that $L^{\otimes k}$ embeds $(X,L)$ into ${\mathbb{P}}^{N_k}$ with image of fixed Hilbert polynomial $P$. The Hilbert scheme ${\mathrm{Hilb}}^P({\mathbb{P}}^{N_k})$ is an algebraic variety (Grothendieck 1961). The locus of polarised pairs inside ${\mathrm{Hilb}}^P$ is an open subscheme. Taking the quotient by the action of $\mathrm{PGL}(N_k+1)$ (the change of basis of $H^0(L^{\otimes k})$) gives the moduli stack $\mathcal{M}$ as an algebraic quotient stack (Mumford 1965, GIT). The result is an Artin algebraic stack — algebraic in the precise sense of derived algebraic geometry. $\square$

Proposition 5 (Hodge-class formulation). A compact Kähler manifold $X$ is projective iff there exists $\alpha \in H^2(X;\mathbb{Z})$ such that $\alpha$ is of type $(1,1)$ and represents a positive $(1,1)$-form in $H^{1,1}(X;\mathbb{R})$.

Proof. By the Lefschetz $(1,1)$-theorem, $\alpha$ Hodge of type $(1,1)$ $\iff$ $\alpha =c_1(L)$ for some holomorphic line bundle $L$. The condition that $\alpha$ admits a positive $(1,1)$-form representative is the same as $L$ being positive (in the Hermitian-metric sense). By Proposition 2, $X$ projective $\iff$ positive line bundle exists $\iff$ positive Hodge class exists. $\square$

Connections [Master]

• Kodaira vanishing 04.09.02. The direct prerequisite: the proof of Kodaira embedding factors entirely through repeated application of Kodaira vanishing to $L^{\otimes k}\otimes {\omega }_X^{-1}$ tensored with ideal sheaves ${\mathcal{I}}_{p,q}$ and ${\mathfrak{m}}_p^2$. Without Kodaira vanishing, the separation-of-points and separation-of-tangents lemmas fail. The Akizuki-Nakano refinement to ${\Omega }^p$-valued cohomology gives sharper bounds on the embedding degree $k_0$.

• Hodge decomposition 04.09.01. The Hodge-class formulation of Kodaira embedding (K3) uses the Hodge decomposition to identify the kernel of the exponential connecting map $\delta :H^2(X;\mathbb{Z})\to H^2(X;{\mathcal{O}}_X)=H^{0,2}$ with integral classes of type $(1,1)$. The local Torelli theorem for K3 surfaces — central to the worked example — is a statement about how the Hodge decomposition determines the complex structure.

• Ample line bundle 04.05.05. In algebraic geometry, an ample line bundle on a projective variety is defined cohomologically (high tensor powers separate points and tangents). Kodaira embedding identifies the analytic notion of positive with the algebraic notion of ample: the analytic positivity criterion (Chern curvature positive) is equivalent, on smooth projective varieties, to ampleness. This is part of the analytic-algebraic dictionary that GAGA (Serre 1956) and Kodaira embedding together make rigorous.

• Serre duality 04.08.03. Used in the Riemann-Roch computation of $h^0(L^{\otimes k})$: for $k$ large, $h^i(L^{\otimes k})=0$ for $i>0$ by Kodaira, so $h^0=\chi$, and the asymptotic growth $\chi (L^{\otimes k})\sim L^n{k}^n/n!$ from Hirzebruch-Riemann-Roch gives the dimension of the embedding target. Serre duality is the source of $h^n(L^{\otimes k}{)}^{\vee }\cong h^0({\omega }_X\otimes L^{\otimes -k})$ and parallel computations.

• Moduli of curves 04.10.01. The polarised-moduli consequence is exemplified by ${\mathcal{M}}_g$: a smooth projective curve of genus $g\ge 2$ carries a canonical polarisation by ${\omega }_X$ (positive since $\mathrm{deg}{\omega }_X=2g-2\ge 2$), and the moduli is automatically algebraic by Kodaira embedding applied uniformly across families. The Deligne-Mumford compactification ${\overline{\mathcal{M}}}_g$ extends to stable curves.

• ddbar-lemma 04.09.05. The ddbar-lemma is part of the analytic toolkit supporting the proof of Kodaira embedding: the canonical Hodge filtration on $H^2$ (made well-defined by ddbar) underlies the identification of positive integral $(1,1)$-classes with positive line bundles, and ddbar ensures the resulting Chern-curvature positivity statement is metric-independent at the level of cohomology classes.

• Hard Lefschetz theorem 04.09.07. Hard Lefschetz organises the cohomology of the embedded projective variety into the Lefschetz primitive decomposition, and the positive integral $(1,1)$-class supplying the Kodaira embedding is exactly the Kähler class whose iterated powers $L^k$ realise the Lefschetz isomorphisms. The two theorems together produce the structural picture of cohomology on a smooth complex projective variety.

• Hodge-Riemann bilinear relations 04.09.08. The Hodge-Riemann relations supply the polarisation underlying Kodaira embedding: a positive integral $(1,1)$-class polarises the weight-2 Hodge structure by HR2, and the resulting positivity is the analytic input that produces enough global sections of high tensor powers to give a projective embedding. The embedding criterion is essentially the existence of an HR2-polarising class in $H^2(X,\mathbb{Z})$.

• Lefschetz (1,1)-theorem 04.09.09. The Lefschetz (1,1)-theorem identifies integral $(1,1)$-classes with line bundles, and Kodaira embedding adds the positivity refinement: a positive integral $(1,1)$-class corresponds to an ample line bundle, which produces a projective embedding for high tensor powers. Together they form the analytic-to-algebraic dictionary positive $(1,1)$-class $\iff$ ample line bundle $\iff$ projective embedding.

• Akizuki-Nakano vanishing theorem 04.09.10. Akizuki-Nakano supplies the sharper vanishing input that Kodaira embedding uses to control the tangent-separation step in the proof: the $p=1$ case $H^q(X,{\Omega }_X^1\otimes L^{\otimes k})=0$ for $q\ge 1$ and $k\gg 0$ ensures that sections of $L^{\otimes k}$ separate not just points but also tangent directions, completing the closed-embedding criterion.

• Toric degeneration of a Calabi-Yau variety 04.12.07. The smooth Calabi-Yau fibres ${\mathfrak{X}}_t$ of a toric degeneration in [04.12.07] are projective varieties, embedded in ${\mathbb{P}}^N$ via the Kodaira embedding theorem of this unit applied to a sufficiently positive line bundle. For Calabi-Yau hypersurfaces in toric Fano varieties, the polarisation comes from the ambient toric polarisation $L_P$ of the polytope-fan dictionary; Kodaira embedding guarantees that the resulting line bundle is sufficiently positive to give a projective embedding. The toric-degeneration setup respects the polarisation throughout the family: the central fibre ${\mathfrak{X}}_0$ is itself projectively embedded via the same lattice-point basis, with toric components glued along their projective sub-embeddings.

• Period integral and the mirror map (pointer) 04.12.13. Kodaira embedding ensures that the Calabi-Yau fibres of the Gross-Siebert reconstructed family are projective and polarised, so that the period integrals of [04.12.13] are genuinely algebraic-geometric objects rather than purely analytic ones. The strictly convex piecewise-affine polarisation $\phi$ on the polarised tropical manifold corresponds, on the algebraic-geometric side, to the relative ample polarisation supplied by Kodaira embedding on each fibre; the period integral apparatus reads off the variation of Hodge structure on this polarised projective family.

Historical & philosophical context [Master]

Kunihiko Kodaira's 1954 paper On Kähler varieties of restricted type (Annals of Mathematics 60, 28–48) introduced what is now called the Kodaira embedding theorem ^{[Kodaira 1954]}. The setting was Hodge varieties — compact Kähler manifolds equipped with a positive integral $(1,1)$-class. Kodaira's terminology "varieties of restricted type" referred to the integrality restriction on the Kähler class; the modern terminology Hodge manifold or polarised Kähler manifold came later.

Kodaira's proof was based on the harmonic-form methods he had developed in the 1953 paper on the Kodaira vanishing theorem ^{[Kodaira 1953]} 04.09.02. The 1954 paper used the vanishing to prove the separation-of-points and separation-of-tangents statements, applying the Hörmander-style $\bar{\partial }$-techniques: a positive line bundle $L$ has positive Chern curvature, which generates a positive Bochner-Kodaira-Nakano inequality, which kills $\bar{\partial }$-harmonic forms of higher degree, which makes the relevant restriction maps surjective.

The full story has multiple historical strands.

The pre-history (1940s). Hodge's 1941 monograph The Theory and Applications of Harmonic Integrals ^{[Hodge 1941]} set up harmonic-integral methods. Chow's 1949 paper On compact complex analytic varieties (Amer. J. Math. 71, 893–914) ^{[Chow 1949]} proved that closed analytic subspaces of ${\mathbb{P}}^N$ are algebraic — the conversion from analytic embedding to algebraic structure. Kodaira and Spencer's deformation theory (1958, three foundational papers in Annals of Math.) provided the moduli-theoretic context.

The Kodaira papers (1953-54). Kodaira-1953 (PNAS) introduced vanishing; Kodaira-1954 (Annals) extended to the embedding theorem. Both papers used harmonic-form analysis as the primary tool. The proofs were transcendental — relying on PDE methods and complex differential geometry. The algebraic content (projective variety, ample line bundle) was a consequence, not a starting point.

The algebraic re-formulation (1960s). Serre's GAGA paper Géométrie algébrique et géométrie analytique (Ann. Inst. Fourier 6, 1956) gave the analytic-algebraic equivalence for projective varieties. Grothendieck's EGA (1960-67) reformulated algebraic geometry on schemes, identifying ample line bundles cohomologically (via the asymptotic Serre theorem $H^i(X;L^{\otimes k})=0$ for $i>0$ and $k\gg 0$). The analytic notion of positivity and the algebraic notion of ampleness became interchangeable through Kodaira embedding.

The polarised-moduli framework (1960s-80s). Mumford's Geometric Invariant Theory (1965) ^{[Kodaira 1954]} developed quotient constructions for moduli of polarised varieties. The Hilbert scheme (Grothendieck 1961) and the GIT quotient (Mumford 1965) together built moduli spaces of polarised varieties as algebraic objects. By the mid-1980s, moduli of polarised K3 surfaces, polarised abelian varieties, and polarised Calabi-Yau manifolds were standard constructions in algebraic geometry.

Moishezon's theorem (1966). Moishezon's On n-dimensional compact complex varieties with n algebraically independent meromorphic functions (Izv. Akad. Nauk SSSR 30, 133–174) ^{[Moishezon 1966]} characterised Kähler-Moishezon manifolds as projective. The result clarified the gap between Kähler and projective: Moishezon non-Kähler examples (Hironaka 1960) show the Kähler hypothesis is essential in Kodaira embedding.

Non-Kähler obstructions. The Hopf surface and related compact complex surfaces of class VII (Bogomolov, Inoue) showed that "Kähler" is a strong constraint on compact complex surfaces. The classification of compact complex surfaces (Kodaira 1964-68) sorted them into Kähler (and hence accessible to Kodaira embedding when polarised) and non-Kähler (where the theorem does not apply).

Modern extensions. Demailly's Complex Analytic and Differential Geometry (online manuscript, 1997-present) ^[Demailly] systematised the positivity-vanishing-embedding framework using $L^2$-methods and multiplier-ideal sheaves. The Nadel vanishing theorem (1990) extended Kodaira to singular metrics, applied to non-smooth varieties. The work of Birkar-Cascini-Hacon-McKernan (2010) on the existence of minimal models for varieties of log general type used Kodaira-vanishing-style arguments throughout.

Kodaira embedding has become one of the foundational theorems of complex geometry — alongside Hodge decomposition, Kodaira vanishing, and Serre duality, it forms the analytic-algebraic backbone of the subject. Kodaira was awarded the Fields Medal in 1954, partly for the embedding theorem and the surface classification, partly for the vanishing theorem and the harmonic-integral methods. The 1954 Annals paper remains the standard reference, with Voisin's Hodge Theory §7.4 ^[Voisin] and Griffiths-Harris §1.4 ^{[Griffiths-Harris]} providing modern expositions.

Bibliography [Master]

@article{Kodaira1954,
  author = {Kodaira, Kunihiko},
  title = {On {K}\"ahler varieties of restricted type (an intrinsic characterization of algebraic varieties)},
  journal = {Annals of Mathematics},
  volume = {60},
  year = {1954},
  pages = {28--48},
}

@article{Kodaira1953,
  author = {Kodaira, Kunihiko},
  title = {On a differential-geometric method in the theory of analytic stacks},
  journal = {Proc. Nat. Acad. Sci. USA},
  volume = {39},
  year = {1953},
  pages = {1268--1273},
}

@book{Hodge1941,
  author = {Hodge, W. V. D.},
  title = {The Theory and Applications of Harmonic Integrals},
  publisher = {Cambridge University Press},
  year = {1941},
  edition = {2nd ed. 1952},
}

@article{Chow1949,
  author = {Chow, Wei-Liang},
  title = {On compact complex analytic varieties},
  journal = {American Journal of Mathematics},
  volume = {71},
  year = {1949},
  pages = {893--914},
}

@article{Moishezon1966,
  author = {Moishezon, Boris},
  title = {On $n$-dimensional compact complex varieties with $n$ algebraically independent meromorphic functions},
  journal = {Izv. Akad. Nauk SSSR Ser. Mat.},
  volume = {30},
  year = {1966},
  pages = {133--174, 345--386, 621--656},
}

@book{Voisin2002,
  author = {Voisin, Claire},
  title = {Hodge Theory and Complex Algebraic Geometry, I},
  publisher = {Cambridge University Press},
  year = {2002},
}

@book{GriffithsHarris1978,
  author = {Griffiths, Phillip and Harris, Joseph},
  title = {Principles of Algebraic Geometry},
  publisher = {Wiley-Interscience},
  year = {1978},
}

@book{Wells1980,
  author = {Wells, Raymond O.},
  title = {Differential Analysis on Complex Manifolds},
  publisher = {Springer},
  series = {Graduate Texts in Mathematics},
  volume = {65},
  year = {1980},
}

@misc{Demailly,
  author = {Demailly, Jean-Pierre},
  title = {Complex Analytic and Differential Geometry},
  note = {Online manuscript, Universit\'e Grenoble Alpes, 1997-present},
}

@book{Mumford1965,
  author = {Mumford, David},
  title = {Geometric Invariant Theory},
  publisher = {Springer},
  series = {Ergebnisse der Mathematik und ihrer Grenzgebiete},
  year = {1965},
}

@article{Nadel1990,
  author = {Nadel, Alan},
  title = {Multiplier ideal sheaves and {K}\"ahler-{E}instein metrics of positive scalar curvature},
  journal = {Annals of Mathematics},
  volume = {132},
  year = {1990},
  pages = {549--596},
}

@article{BCHM2010,
  author = {Birkar, Caucher and Cascini, Paolo and Hacon, Christopher and McKernan, James},
  title = {Existence of minimal models for varieties of log general type},
  journal = {Journal of the American Mathematical Society},
  volume = {23},
  year = {2010},
  pages = {405--468},
}