Serre's GAGA comparison theorem

Intuition Beginner

Take a shape cut out by polynomial equations inside complex projective space — say a curve $\{F=0\}$ in ${\mathbb{CP}}^2$. You can study it two ways. The algebraic way uses only polynomials and rational functions. The analytic way lets you use every convergent power series, every holomorphic function. The analytic toolkit looks far richer. GAGA says that on projective space the extra power buys you nothing new.

GAGA is short for Géométrie Algébrique et Géométrie Analytique. Jean-Pierre Serre proved it in 1956. The headline: on a projective complex variety, "holomorphic" and "algebraic" secretly name the same objects. Every holomorphic vector bundle is an algebraic one. Every holomorphic map between two such varieties is given by polynomials. Every analytic subset is the zero set of polynomials.

The catch is the word projective. Projective space is compact: closed and bounded with no edges running off to infinity. Compactness is what forces holomorphic things to be algebraic. Drop it and the bridge breaks. On the open plane the function $e^z$ is holomorphic but is no polynomial, so on non-compact spaces the two worlds genuinely differ.

So GAGA is a bridge. One the algebraic half of the corpus and the complex-analytic half walk across constantly.

Visual Beginner

A projective variety viewed twice — once through the algebraic lens of polynomials and once through the analytic lens of holomorphic functions — with an arrow showing the two pictures coincide.

Worked example Beginner

Consider the simplest projective variety: a single point of ${\mathbb{CP}}^1$, or all of ${\mathbb{CP}}^1$ itself.

A global holomorphic function on the compact Riemann sphere ${\mathbb{CP}}^1$ must be constant. The maximum-modulus principle pins it down: a holomorphic function on a compact connected complex manifold attains its maximum, so it cannot vary. On the algebraic side, the global regular functions on ${\mathbb{CP}}^1$ are also just the constants $\mathbb{C}$. The two answers match.

Now compare line bundles. The holomorphic line bundles on ${\mathbb{CP}}^1$ are $\mathcal{O}(n)$ for $n\in \mathbb{Z}$, each an integer's worth of twisting. The algebraic line bundles on ${\mathbb{P}}^1$ are also $\mathcal{O}(n)$ for $n\in \mathbb{Z}$. Same list. GAGA is the structural reason these two classifications agree: the dictionary between holomorphic and algebraic bundles on a projective variety is exact.

Contrast the open disc $\{|z|<1\}$, which is not projective. There the global holomorphic functions form a vast space — every convergent power series — while the regular functions are far fewer. No GAGA there. Compactness, supplied by projective space, is the missing ingredient that makes the worked example come out even.

Check your understanding Beginner

Formal definition Intermediate+

Let $X$ be a scheme of finite type over $\mathbb{C}$. Its set of closed points carries a canonical structure of a complex-analytic space $X^{\mathrm{an}}$: locally $X$ is cut out of ${\mathbb{A}}_{\mathbb{C}}^n$ by polynomials $f_1,\dots ,f_r$, and $X^{\mathrm{an}}$ is the same vanishing locus regarded as an analytic subspace of ${\mathbb{C}}^n$, with structure sheaf ${\mathcal{O}}_{X^{\mathrm{an}}}$ the sheaf of holomorphic functions modulo the ideal generated by the $f_i$. There is a morphism of locally ringed spaces $\phi :(X^{\mathrm{an}},{\mathcal{O}}_{X^{\mathrm{an}}})\to (X,{\mathcal{O}}_X)$.

Analytification of sheaves. For a coherent ${\mathcal{O}}_X$-module $\mathcal{F}$, define its analytification

$${\mathcal{F}}^{\mathrm{an}}:={\phi }^{\ast }\mathcal{F}={\mathcal{O}}_{X^{\mathrm{an}}}{\otimes }_{{\phi }^{-1}{\mathcal{O}}_X}{\phi }^{-1}\mathcal{F}.$$

By Oka's coherence theorem ${\mathcal{O}}_{X^{\mathrm{an}}}$ is a coherent analytic sheaf, so ${\mathcal{F}}^{\mathrm{an}}$ is a coherent analytic sheaf on $X^{\mathrm{an}}$. This produces an exact functor

$$(-{)}^{\mathrm{an}}:\mathrm{Coh}(X)⟶\mathrm{Coh}(X^{\mathrm{an}}).$$

Theorem (GAGA; Serre 1956). Let $X$ be a projective scheme over $\mathbb{C}$. Then:

(G1) For every coherent ${\mathcal{O}}_X$-module $\mathcal{F}$ and every $i\ge 0$, the natural comparison map is an isomorphism

$$H^i(X,\mathcal{F})\stackrel{\sim }{\to }{H}^i(X^{\mathrm{an}},{\mathcal{F}}^{\mathrm{an}}).$$

(G2) The functor $(-{)}^{\mathrm{an}}:\mathrm{Coh}(X)\to \mathrm{Coh}(X^{\mathrm{an}})$ is an equivalence of categories: it is fully faithful, and every coherent analytic sheaf on $X^{\mathrm{an}}$ is isomorphic to ${\mathcal{G}}^{\mathrm{an}}$ for a unique-up-to-isomorphism coherent algebraic $\mathcal{G}$.

Full faithfulness, explicitly. For coherent $\mathcal{F},\mathcal{G}$ on $X$ the analytification of $\mathrm{Hom}$ is a bijection

$${\mathrm{Hom}}_{{\mathcal{O}}_X}(\mathcal{F},\mathcal{G})\stackrel{\sim }{\to }{\mathrm{Hom}}_{{\mathcal{O}}_{X^{\mathrm{an}}}}({\mathcal{F}}^{\mathrm{an}},{\mathcal{G}}^{\mathrm{an}}).$$

This follows from (G1) applied to the coherent sheaf $\mathcal{H}om(\mathcal{F},\mathcal{G})$ in degree $i=0$.

Consequences (stated; proved in later sections).

(C1) Chow's theorem. Every closed analytic subvariety of ${\mathbb{CP}}^n$ is algebraic — the zero locus of finitely many homogeneous polynomials.

(C2) Holomorphic maps are morphisms. Every holomorphic map $f:X\to Y$ between projective complex varieties is a morphism of varieties (given by polynomials in homogeneous coordinates).

(C3) Bundles are algebraic. Every holomorphic vector bundle on a projective variety is the analytification of an algebraic one; the two Picard groups agree, $\mathrm{Pic}(X)\cong \mathrm{Pic}(X^{\mathrm{an}})$.

The role of projectivity. Each of (G1), (G2) uses properness in an essential way (projective $\Rightarrow$ proper). On a non-proper variety the statement is false. The affine line ${\mathbb{A}}_{\mathbb{C}}^1$ has $X^{\mathrm{an}}=\mathbb{C}$, and $e^z\in H^0(\mathbb{C},{\mathcal{O}}_{\mathbb{C}}^{\mathrm{an}})$ is a global analytic section with no algebraic counterpart: the comparison map $H^0({\mathbb{A}}^1,\mathcal{O})=\mathbb{C}[z]\hookrightarrow H^0(\mathbb{C},{\mathcal{O}}^{\mathrm{an}})$ is injective but very far from surjective.

Key theorem with proof Intermediate+

Theorem (GAGA cohomology comparison). Let $X\subseteq {\mathbb{P}}_{\mathbb{C}}^N$ be a projective scheme and $\mathcal{F}$ a coherent ${\mathcal{O}}_X$-module. Then $H^i(X,\mathcal{F})\cong H^i(X^{\mathrm{an}},{\mathcal{F}}^{\mathrm{an}})$ for all $i\ge 0$, and $(-{)}^{\mathrm{an}}$ is an equivalence onto $\mathrm{Coh}(X^{\mathrm{an}})$.

Proof outline. It suffices to treat $X={\mathbb{P}}^N$, since a coherent sheaf on a closed subscheme is one on ${\mathbb{P}}^N$ pushed forward, and both sides commute with the closed embedding.

Step 1 — anchor on the twisting sheaves. First establish the comparison for $\mathcal{F}={\mathcal{O}}_{{\mathbb{P}}^N}(n)$, every $n\in \mathbb{Z}$. Both algebraic and analytic cohomologies of $\mathcal{O}(n)$ on ${\mathbb{P}}^N$ are computed explicitly and agree dimension by dimension: $H^0$ is the space of degree-$n$ homogeneous polynomials on both sides, the top cohomology $H^N$ matches by Serre duality against $\mathcal{O}(-n-N-1)$, and the intermediate groups vanish on both sides. So (G1) holds for each $\mathcal{O}(n)$.

Step 2 — Serre vanishing on both sides. By 04.03.05, for any coherent $\mathcal{F}$ there is $n_0$ with $H^i(X,\mathcal{F}(n))=0$ for $i>0$ and $n\ge n_0$, and $\mathcal{F}(n)$ globally generated. The analytic analogue — Cartan's Theorem B and the finiteness theorem for coherent sheaves on the compact space $X^{\mathrm{an}}$ — gives the matching analytic vanishing. Compactness of $X^{\mathrm{an}}$ is what makes the analytic cohomology finite-dimensional; this is the foundational reason properness cannot be dropped.

Step 3 — dévissage by induction on $i$, descending. Choose a surjection $\mathcal{O}(-n{)}^{\oplus k}\twoheadrightarrow \mathcal{F}$ with kernel $\mathcal{K}$ coherent, giving a short exact sequence $0\to \mathcal{K}\to \mathcal{O}(-n{)}^{\oplus k}\to \mathcal{F}\to 0$. Analytification is exact, so it produces the analytic short exact sequence. The two long exact cohomology sequences map to each other; the comparison maps form a ladder.

Step 4 — five-lemma. On the $\mathcal{O}(-n)$ terms the comparison is an isomorphism by Step 1. Running the long-exact-sequence ladder and applying the five-lemma propagates the isomorphism: knowing the comparison in degree $i+1$ for $\mathcal{K}$ forces it in degree $i$ for $\mathcal{F}$. Descending induction starting from the vanishing of high cohomology (Step 2) closes the loop and proves (G1) for every coherent $\mathcal{F}$.

Step 5 — equivalence of categories. Full faithfulness is (G1) in degree $0$ for $\mathcal{H}om$. Essential surjectivity — every coherent analytic sheaf is algebraic — uses Step 2 again: a coherent analytic $\mathcal{G}$ on the compact $X^{\mathrm{an}}$ has $\mathcal{G}(n)$ globally generated for $n\gg 0$, yielding an analytic presentation $\mathcal{O}(-n{)}^{\oplus a}\to \mathcal{O}(-m{)}^{\oplus b}\to \mathcal{G}\to 0$; the two bundle maps are algebraic by full faithfulness, so $\mathcal{G}$ is the analytification of the algebraic cokernel. $\square$

Bridge. The anchor-and-dévissage shape here builds toward every place the corpus silently crosses between the algebraic and analytic worlds, and the five-lemma ladder appears again in the proofs of Chow's theorem and the comparison of Picard groups below. The foundational reason the argument runs is Serre vanishing on the compact analytic space; this is exactly the same finiteness that powers the algebraic 04.03.05, now read on $X^{\mathrm{an}}$. The bridge is that GAGA generalises the one-line maximum-modulus fact "holomorphic functions on a compact manifold are constant" into a full equivalence of sheaf categories, and putting these together shows why the Hodge-theoretic 04.09.01 and complex-analytic 06.10.22 units may freely import algebraic input.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none — Mathlib has projective schemes and coherent sheaves on the algebraic side, but no complex-analytic spaces, no holomorphic structure sheaf, no analytification functor, and no coherent-analytic-sheaf category, so the comparison statement cannot yet even be phrased.

import Mathlib.AlgebraicGeometry.ProjectiveSpectrum.Scheme

namespace Codex.AlgebraicGeometry.Hodge

-- GAGA (Serre 1956): for a projective scheme X over ℂ, the analytification
-- functor Coh(X) → Coh(Xᵃⁿ) is an equivalence of categories inducing
-- isomorphisms Hⁱ(X, F) ≅ Hⁱ(Xᵃⁿ, Fᵃⁿ) for all i.
-- Blocked: no complex-analytic spaces, no Oka coherence, no analytification
-- functor, no comparison-of-sites cohomology in Mathlib.

end Codex.AlgebraicGeometry.Hodge

Advanced results Master

Relative GAGA (Grothendieck). Serre's theorem was recast and generalised by Grothendieck (SGA 1, Exp. XII) into a relative statement: for a proper morphism $f:X\to S$ of schemes locally of finite type over $\mathbb{C}$, analytification commutes with higher direct images, $(R^i{f}_{\ast }\mathcal{F}{)}^{\mathrm{an}}\cong R^i{f}_{\ast }^{\mathrm{an}}({\mathcal{F}}^{\mathrm{an}})$, and the relative analytification functor is an equivalence on coherent sheaves over each fibre. The absolute GAGA is the case $S=\mathrm{Spec}\mathbb{C}$. Properness, not projectivity, is the true hypothesis: Grothendieck's proof uses Chow's lemma to reduce the proper case to the projective case where Serre's argument applies.

The étale fundamental group comparison. A parallel comparison (SGA 1, Exp. XII, the "Riemann existence theorem" in modern dress) states that for a variety $X$ over $\mathbb{C}$ the finite étale covers of $X$ correspond to finite topological covers of $X^{\mathrm{an}}$, giving ${\pi }_1^{\acute{\mathrm{e}}\mathrm{t}}(X)\cong \widehat{{\pi }_1(X^{\mathrm{an}})}$, the profinite completion of the topological fundamental group. This is the GAGA principle applied to covers rather than to coherent sheaves, and it is the foundation of the comparison between algebraic and topological monodromy.

Failure quantified: the affine counterexample. On ${\mathbb{A}}_{\mathbb{C}}^1$ the comparison map $H^0({\mathbb{A}}^1,\mathcal{O})=\mathbb{C}[z]\hookrightarrow H^0(\mathbb{C},{\mathcal{O}}^{\mathrm{an}})=\{\text{entire functions}\}$ has image the polynomials, a countable-dimensional subspace of an uncountable-dimensional target; $e^z$, $\sin z$, $\sum z^n/n!$ all live outside. The failure is total, not marginal — exactly the gap properness closes.

Algebraisation of formal and rigid analytic geometry. GAGA has non-archimedean siblings: Köpf's rigid GAGA and Berthelot's, and the formal-GAGA of Grothendieck's EGA III §5, which compares coherent sheaves on a proper scheme with sheaves on its formal completion. These share the architecture: a proper morphism, a comparison of sites, a twisting-plus-dévissage proof. The pattern is one of the deepest recurring devices in algebraic geometry.

Hodge-theoretic payoff. GAGA is the silent hinge that lets the Hodge decomposition 04.09.01 be stated for projective varieties using holomorphic differential forms while computing with algebraic coherent cohomology $H^q(X,{\Omega }_X^p)$. Without GAGA the identification of analytic Dolbeault cohomology with algebraic sheaf cohomology of the algebraic de Rham complex would be a separate theorem; GAGA supplies it for free on projective $X$.

Synthesis. GAGA is the bridge that makes the algebraic and analytic halves of the corpus into one theory: putting these together, the foundational reason every projective Hodge-theoretic computation may be run algebraically is that the analytification functor is an equivalence of coherent categories inducing isomorphisms on all cohomology. This is exactly the statement that "holomorphic equals algebraic on projective space," and it generalises in two directions at once — Grothendieck's relative and formal versions show the true hypothesis is properness, while the étale comparison shows the same principle is dual to the comparison of fundamental groups. The central insight, recurring from Chow's theorem through rigid GAGA, is that compactness of the analytic space forces finiteness and hence algebraicity; this pattern recurs wherever the corpus crosses from 04.06.02 coherent algebra to 06.10.22 complex-analytic coherence, and the bridge is load-bearing for 04.09.01, 04.09.11 and the transcendental chapter as a whole.

Full proof set Master

Proposition (Serre vanishing transfers across analytification). Let $X\subseteq {\mathbb{P}}_{\mathbb{C}}^N$ be projective and $\mathcal{F}$ coherent on $X$. There is $n_0$ such that for all $n\ge n_0$ and all $i>0$,

$$H^i(X,\mathcal{F}(n))=0\text{and}{H}^i(X^{\mathrm{an}},{\mathcal{F}}^{\mathrm{an}}(n))=0,$$

and $\mathcal{F}(n)$, ${\mathcal{F}}^{\mathrm{an}}(n)$ are globally generated.

Proof. The algebraic half is Serre's vanishing and finiteness theorem 04.03.05: ampleness of ${\mathcal{O}}_X(1)$ gives $n_0$ killing higher cohomology and forcing global generation. For the analytic half, $X^{\mathrm{an}}$ is a compact complex space (projective $\Rightarrow$ proper $\Rightarrow$ compact analytification). On a compact complex space the coherent cohomology of any coherent analytic sheaf is finite-dimensional (Cartan-Serre finiteness), and the ample line bundle $\mathcal{O}(1{)}^{\mathrm{an}}$ is positive, so the Kodaira-type analytic vanishing theorem (or Cartan's Theorem A/B applied after the twisting) provides $n_1$ with $H^i(X^{\mathrm{an}},{\mathcal{F}}^{\mathrm{an}}(n))=0$ for $i>0,n\ge n_1$ and global generation by Theorem A. Take $n_0=\max (n_0^{\mathrm{alg}},n_1)$. Finiteness on the compact analytic side is the property that fails on non-compact $X^{\mathrm{an}}$ and is the load-bearing input to the dévissage. $\square$

Proposition (full faithfulness of analytification). For coherent $\mathcal{F},\mathcal{G}$ on a projective $X$, the map ${\mathrm{Hom}}_X(\mathcal{F},\mathcal{G})\to {\mathrm{Hom}}_{X^{\mathrm{an}}}({\mathcal{F}}^{\mathrm{an}},{\mathcal{G}}^{\mathrm{an}})$ is a bijection.

Proof. Homomorphisms of coherent sheaves are global sections of the coherent sheaf $\mathcal{H}om(\mathcal{F},\mathcal{G})$: ${\mathrm{Hom}}_X(\mathcal{F},\mathcal{G})=H^0(X,\mathcal{H}om(\mathcal{F},\mathcal{G}))$, and analytification commutes with $\mathcal{H}om$ for coherent modules, $\mathcal{H}om(\mathcal{F},\mathcal{G}{)}^{\mathrm{an}}\cong \mathcal{H}om({\mathcal{F}}^{\mathrm{an}},{\mathcal{G}}^{\mathrm{an}})$. The cohomology comparison (G1) in degree $i=0$ applied to the coherent sheaf $\mathcal{H}om(\mathcal{F},\mathcal{G})$ yields the bijection $H^0(X,\mathcal{H}om)\cong H^0(X^{\mathrm{an}},\mathcal{H}o{m}^{\mathrm{an}})$, which is the asserted isomorphism on Hom-sets. $\square$

The degree-$0$ comparison feeds full faithfulness; the descending induction with the five-lemma upgrades it to all degrees and, with the Proposition above supplying the vanishing anchor, to essential surjectivity — completing the equivalence proved in the Key theorem section.

Connections Master

• Coherent sheaf 04.06.02 — GAGA is a statement about the category $\mathrm{Coh}(X)$ and its analytic analogue; the analytification functor is defined precisely on coherent modules, and coherence on both sides (algebraic finiteness, Oka coherence) is the hypothesis that makes the comparison run.

• Serre vanishing and finiteness 04.03.05 — the algebraic vanishing $H^i(\mathcal{F}(n))=0$ for $n\gg 0$ is the anchor of the dévissage; GAGA pairs it with its analytic twin on the compact space $X^{\mathrm{an}}$ to drive the five-lemma induction.

• Projective scheme 04.02.03 — projectivity (hence properness, hence compactness of $X^{\mathrm{an}}$) is the indispensable hypothesis; GAGA fails on affine and non-proper schemes, the affine line with $e^z$ being the standard counterexample.

• Hodge decomposition 04.09.01 — GAGA is the silent bridge letting the Hodge decomposition $H^k=⨁H^q(X,{\Omega }_X^p)$ be computed with algebraic coherent cohomology while interpreted analytically via holomorphic forms on $X^{\mathrm{an}}$.

• Complex spaces and coherence 06.10.22 — supplies the analytic half: coherent analytic sheaves, Oka coherence of the holomorphic structure sheaf, and Cartan-Serre finiteness on compact complex spaces, which are the objects GAGA matches to algebraic coherent sheaves.

• Kodaira embedding theorem 04.09.11 — together GAGA and Kodaira close the loop: Kodaira embeds a positive compact Kähler manifold into ${\mathbb{P}}^N$ analytically, and GAGA then certifies that the resulting analytic submanifold is an algebraic projective variety, so "positive Kähler" implies "projective algebraic."

Historical & philosophical context Master

Jean-Pierre Serre's Géométrie algébrique et géométrie analytique ^{[Serre 1956]} appeared in the Annales de l'Institut Fourier in 1956, three years after his foundational Faisceaux algébriques cohérents (FAC) had recast algebraic geometry in the language of coherent sheaves and their cohomology. GAGA was the natural sequel: having built coherent-sheaf cohomology algebraically, Serre asked precisely how it compared to the older complex-analytic cohomology of Cartan and Oka. The answer — a clean equivalence of categories on projective varieties — unified two research traditions that had developed largely in parallel: the function-theoretic study of complex manifolds (Riemann, Poincaré, Hodge, Cartan, Oka) and the polynomial-algebraic study of varieties (Zariski, Weil, Serre, Grothendieck).

The result also vindicated and generalised W.-L. Chow's 1949 theorem ^{[Chow 1949]} that a compact analytic subvariety of projective space is algebraic. Chow's proof was direct and geometric; Serre's recovered it as a one-line corollary of a structural statement about sheaves, an instance of the twentieth-century movement to replace ad-hoc geometric arguments with categorical and cohomological machinery. Grothendieck soon absorbed GAGA into his relative framework (SGA 1, Exp. XII) ^{[Grothendieck 1971]}, where the essential hypothesis was revealed to be properness rather than projectivity, and the comparison was extended to the étale fundamental group via the Riemann existence theorem. Philosophically, GAGA crystallised a principle that recurs throughout modern geometry: on a compact (proper) object, transcendental and algebraic descriptions coincide, and the apparent extra freedom of analysis is an illusion that compactness dispels. Hartshorne's Appendix B ^{[Hartshorne 1977]} made the theorem standard graduate material, and it remains the canonical license for importing analytic methods — Hodge theory, Kähler geometry, vanishing theorems — into the study of projective algebraic varieties.

Bibliography Master

@article{Serre1956GAGA,
  author  = {Serre, Jean-Pierre},
  title   = {G\'eom\'etrie alg\'ebrique et g\'eom\'etrie analytique},
  journal = {Annales de l'Institut Fourier},
  volume  = {6},
  pages   = {1--42},
  year    = {1956}
}

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

@book{Hartshorne1977,
  author    = {Hartshorne, Robin},
  title     = {Algebraic Geometry},
  series    = {Graduate Texts in Mathematics},
  number    = {52},
  publisher = {Springer},
  year      = {1977},
  note      = {Appendix B}
}

@book{GriffithsHarris1978,
  author    = {Griffiths, Phillip and Harris, Joseph},
  title     = {Principles of Algebraic Geometry},
  publisher = {Wiley},
  year      = {1978},
  note      = {Chapter 1}
}

@incollection{GrothendieckSGA1XII,
  author    = {Grothendieck, Alexander},
  title     = {Rev\^etements \'etales et groupe fondamental (SGA 1), Expos\'e XII: G\'eom\'etrie alg\'ebrique et g\'eom\'etrie analytique},
  series    = {Lecture Notes in Mathematics},
  number    = {224},
  publisher = {Springer},
  year      = {1971}
}