04.09.07 · algebraic-geometry / hodge

Hard Lefschetz theorem

shipped3 tiersLean: partial

Anchor (Master): Lefschetz 1924 *L'analysis situs et la géométrie algébrique* (Gauthier-Villars); Hodge 1941 *The Theory and Applications of Harmonic Integrals*; Weil 1958 *Variétés Kählériennes*; Voisin *Hodge Theory and Complex Algebraic Geometry* I §6.2 (analytic proof via the Kähler identities and $\mathfrak{sl}_2$); Griffiths-Harris *Principles of Algebraic Geometry* §0.7 (Hodge-Riemann bilinear relations) and §1.2 (Lefschetz decomposition); Deligne 1968 *Théorème de Lefschetz et critères de dégénérescence de suites spectrales* (*Publ. Math. IHÉS* 35) — algebraic version via the Weil conjectures; Beilinson-Bernstein-Deligne 1982 *Faisceaux pervers* (*Astérisque* 100) — decomposition theorem and the perverse-sheaf generalisation

Intuition [Beginner]

The Hard Lefschetz theorem is a single sharp statement about how the Kähler class — the geometric "size" of a compact Kähler manifold — moves cohomology classes between degrees. On a compact Kähler manifold $X$ of complex dimension $n$ , the cup product with the Kähler class $ω$ is a linear map from the cohomology in degree $k$ to the cohomology in degree $k + 2$ . Iterating this map $k$ times produces a map from the cohomology in degree $n - k$ to the cohomology in degree $n + k$ . Hard Lefschetz says: this iterated map is an isomorphism. Not just injective, not just surjective, but a perfect bijection between the cohomology below the middle degree and the cohomology above it.

Why is the statement striking? Poincaré duality already gives a pairing between $H^{n - k}$ and $H^{n + k}$ , identifying their dimensions. Hard Lefschetz strengthens this in two ways. First, it gives an explicit isomorphism — cup product with $ω^{k}$ — rather than just a duality pairing. Second, this isomorphism realises the dimension equality through a single geometric ingredient: the Kähler form raised to a power. The Kähler condition turns out to be what makes the dimensions match in a controlled way.

The theorem unlocks the Lefschetz decomposition of cohomology into pieces called primitive cohomology. The takeaway: on a compact Kähler manifold, the Kähler class is a generator of an action that organises all of cohomology — an $sl_{2}$ -action — and Hard Lefschetz is the visible signature of that action.

Visual [Beginner]

A schematic of a compact Kähler manifold of complex dimension $n$ with its cohomology groups stacked vertically by degree from $0$ to $2 n$ . Two columns of cohomology groups symmetric about the middle row $H^{n}$ . An arrow labelled $L^{k}$ — cup product with the $k$ th power of the Kähler class — runs from $H^{n - k}$ to $H^{n + k}$ , identifying the two groups as isomorphic. A second panel illustrates the Lefschetz decomposition by labelling each row with its sub-pieces: primitive cohomology classes sit at the start of each row, and Lefschetz powers shift them to higher rows.

The picture captures the structural content: cohomology of a compact Kähler manifold is symmetric about the middle degree, and the Kähler class moves classes upward by two degrees at a time. Hard Lefschetz says that $k$ steps upward from the bottom-symmetric row reach the top-symmetric row bijectively.

Worked example [Beginner]

Verify Hard Lefschetz on the complex projective plane $P^{2}$ and the complex projective space $P^{3}$ by writing out the cohomology in each degree and checking that cup product with the Kähler class delivers the predicted isomorphisms.

Step 1. The complex projective plane $P^{2}$ . Complex dimension $n = 2$ . The cohomology is $H^{0} = C$ , $H^{1} = 0$ , $H^{2} = C$ generated by the hyperplane class $ω = [H]$ , $H^{3} = 0$ , $H^{4} = C$ generated by $ω^{2} = [H]^{2}$ . The Kähler class is $ω$ itself.

Step 2. Check Hard Lefschetz on $P^{2}$ . For $k = 1$ : the iterated cup product is $L^{1} = ω \cup -$ , a map from $H^{2 - 1} = H^{1} = 0$ to $H^{2 + 1} = H^{3} = 0$ . The map between two zero spaces is the zero map, vacuously an isomorphism. For $k = 2$ : $L^{2} = ω^{2} \cup -$ , a map from $H^{0} = C$ to $H^{4} = C$ , sending $1$ to $ω^{2} = [H]^{2}$ . This map is an isomorphism since $ω^{2}$ is a non-zero generator of $H^{4}$ . Both predicted iso maps land.

Step 3. The complex projective space $P^{3}$ . Complex dimension $n = 3$ . The cohomology is $H^{2 k} = C$ generated by $ω^{k}$ for $0 \leq k \leq 3$ , all odd-degree cohomology is zero. Hard Lefschetz predicts: $L^{1} : H^{2} \to H^{4}$ sends $ω \to ω^{2}$ , an iso; $L^{2} : H^{1} \to H^{5}$ is $0 \to 0$ , vacuous; $L^{3} : H^{0} \to H^{6}$ sends $1 \to ω^{3}$ , an iso. All checks confirm.

Step 4. The Lefschetz decomposition on $P^{n}$ . On complex projective $n$ -space, every cohomology class is a power of $ω$ , and primitive cohomology $P^{2 k} = ker (L^{n - k + 1})$ vanishes for $k \geq 1$ . The only non-vanishing primitive class is $1 \in H^{0} = P^{0}$ . The Lefschetz decomposition collapses to a single row of powers of $ω$ . This is the cleanest possible case.

What this tells us: Hard Lefschetz on projective space reduces to the statement that $ω^{n}$ is the fundamental class — a non-zero top-degree volume form — and that powers of $ω$ track the full cohomology ring. On more interesting manifolds (K3 surfaces, abelian varieties, Calabi-Yau threefolds), the same theorem says something genuinely new about how cohomology classes are related by the Kähler class.

Check your understanding [Beginner]

Exercise (easy, multiple choice).

On a compact Kähler manifold $X$ of complex dimension $n$ , the Hard Lefschetz theorem states that cup product with the $k$ -th power of the Kähler class is which kind of map from $H^{n - k}$ to $H^{n + k}$ ?

A. Always the zero map
B. Injective but generally not surjective
C. An isomorphism for every $0 \leq k \leq n$
D. Surjective only for $k = n$

Hint

The theorem singles out the specific iterated cup-product map $L^{k} : H^{n - k} \to H^{n + k}$ and makes a sharp claim about its kernel and cokernel.

Answer

C. The Hard Lefschetz theorem asserts that for every $0 \leq k \leq n$ , the iterated cup-product map $L^{k} : H^{n - k} (X, R) \to H^{n + k} (X, R)$ is an isomorphism. Feedback-correct: yes; the iterated cup product with the Kähler class is the canonical isomorphism realising Poincaré duality on a compact Kähler manifold. Feedback-wrong: A is wrong because the map is non-zero on the top-degree generator; B is wrong because the dimensions agree by Poincaré duality, so injective implies surjective; D is wrong because the iso holds for every $k$ , not only $k = n$ .

Formal definition [Intermediate+]

Let $X$ be a compact Kähler manifold of complex dimension $n$ , equipped with a Kähler form $ω \in Ω^{1, 1} (X)$ representing a class $[ω] \in H^{1, 1} (X, R) \subset H^{2} (X, R)$ . The Lefschetz operator is the linear map

L : H^{k} (X, R) \to H^{k + 2} (X, R), L (α) = [ω] \cup α,

cup product with the Kähler class, raising the degree by two. Its $k$ -fold iterate $L^{k} : H^{j} (X, R) \to H^{j + 2 k} (X, R)$ is cup product with $[ω]^{k} = [ω \land \dots \land ω]$ .

Definition (Hard Lefschetz theorem, isomorphism form). The Hard Lefschetz theorem for the compact Kähler manifold $X$ asserts that for every $0 \leq k \leq n$ , the $k$ -fold iterated Lefschetz operator $$ L^k : H^{n-k}(X, \mathbb{R}) \xrightarrow{\sim} H^{n+k}(X, \mathbb{R}) $$ is an isomorphism. Equivalently, the cup-product map is an iso of complex cohomology $H^{n - k} (X, C) \sim H^{n + k} (X, C)$ and respects the Hodge decomposition: $L^{k}$ sends $H^{p, q}$ isomorphically to $H^{p + k, q + k}$ for $p + q = n - k$ .

Definition (primitive cohomology). The primitive cohomology of degree $ℓ$ is the kernel of the Lefschetz raising power that overshoots middle degree: $$ P^{\ell}(X, \mathbb{R}) := \ker\bigl(L^{n - \ell + 1} : H^{\ell}(X, \mathbb{R}) \to H^{2n - \ell + 2}(X, \mathbb{R})\bigr), $$ defined for $0 \leq ℓ \leq n$ . By convention $P^{ℓ} = 0$ for $ℓ > n$ . The bidegree refinement $P^{p, q} := P^{p + q} \cap H^{p, q}$ is the primitive part of $H^{p, q}$ .

Definition (Lefschetz decomposition). The Hard Lefschetz theorem implies the Lefschetz decomposition of cohomology: for $0 \leq k \leq 2 n$ , $$ H^k(X, \mathbb{R}) = \bigoplus_{j \geq 0,\ k - 2j \geq 0} L^j P^{k - 2j}(X, \mathbb{R}), $$ a direct-sum decomposition expressing $H^{k}$ as a sum of Lefschetz powers acting on primitive cohomology of lower degree.

Definition (the Lefschetz $sl_{2}$ -triple). Let $Λ : H^{k} (X) \to H^{k - 2} (X)$ be the Lefschetz lowering operator, the formal adjoint of $L$ with respect to the $L^{2}$ -inner product on differential forms induced by the Kähler metric. Let $H : H^{k} (X) \to H^{k} (X)$ act as multiplication by $k - n$ — the operator measuring the degree displacement from middle dimension. Then the triple $(L, Λ, H)$ satisfies the bracket relations of $sl_{2}$ : $$ [L, \Lambda] = H, \qquad [H, L] = 2L, \qquad [H, \Lambda] = -2\Lambda, $$ turning $H^{∙} (X, R)$ into a finite-dimensional representation of the Lie algebra $sl_{2} (R)$ .

Counterexamples to common slips

Kähler hypothesis is essential. On a compact complex non-Kähler manifold, the Lefschetz operator $L$ is still defined whenever a $(1, 1)$ -form is chosen, but the bracket relations $[L, Λ] = H$ fail to identify the Riemannian and Dolbeault Laplacians, and the iterated cup-product map $L^{k}$ may fail to be injective or surjective. Hopf surfaces and Iwasawa manifolds furnish explicit counterexamples.
Coefficient choice. Hard Lefschetz over $C$ on a compact Kähler manifold is a statement about $R$ -cohomology that automatically extends by tensoring with $C$ and refines bidegree-wise on the Hodge decomposition. The integral statement $L^{k} : H^{n - k} (X, Z) \to H^{n + k} (X, Z)$ is not an isomorphism in general — torsion classes can obstruct integrality. The clean statement is in $R$ or $Q$ coefficients.
Kähler-class dependence. The Lefschetz operator $L = ω \cup -$ depends on the choice of Kähler class $[ω]$ , but the iso-statement of Hard Lefschetz holds for every Kähler class. Different Kähler classes give different isomorphisms but the same isomorphism property. Hard Lefschetz is a property of the manifold and its Kähler cone, not of a specific Kähler class.
Off-by-one in primitive cohomology. Primitive cohomology is defined as the kernel of $L^{n - ℓ + 1}$ , not $L^{n - ℓ}$ . The off-by-one accounts for the fact that $L^{n - ℓ}$ sends $H^{ℓ}$ to $H^{2 n - ℓ}$ (an iso by Hard Lefschetz), so its kernel is zero; the next power $L^{n - ℓ + 1}$ sends $H^{ℓ}$ to $H^{2 n - ℓ + 2} = 0$ (above top degree), so its kernel is all of $H^{ℓ}$ at the cohomology level. The correct kernel is the one cutting off the highest-weight component of each $sl_{2}$ -irreducible inside $H^{ℓ}$ .

Key theorem with proof [Intermediate+]

Theorem (Hard Lefschetz; Voisin Vol I §6.2). Let $X$ be a compact Kähler manifold of complex dimension $n$ with Kähler class $[ω] \in H^{1, 1} (X, R)$ . For every $0 \leq k \leq n$ , the $k$ -fold cup-product map $$ L^k = [\omega]^k \cup - : H^{n - k}(X, \mathbb{R}) \to H^{n + k}(X, \mathbb{R}) $$ is an isomorphism. The map respects the Hodge decomposition: $L^{k}$ sends $H^{p, q}$ isomorphically to $H^{p + k, q + k}$ for every $p + q = n - k$ .

Proof. The argument has four steps. First, define the Lefschetz adjoint $Λ$ and verify the bracket relations of $sl_{2}$ . Second, observe that the de Rham complex carries a finite-dimensional $sl_{2}$ -representation. Third, decompose this representation into irreducibles by highest weight. Fourth, read Hard Lefschetz off the irreducible structure.

Step 1: the $sl_{2}$ -action. Fix a Kähler metric on $X$ and the associated Kähler form $ω$ . Define the Lefschetz operator on differential forms by $L (α) = ω \land α$ , and its formal adjoint $Λ$ with respect to the $L^{2}$ -inner product $⟨ α, β ⟩ = \int_{X} α \land ⋆ \overset{ˉ}{β}$ induced by the Kähler metric. The grading operator $H$ acts on a $k$ -form as multiplication by $k - n$ . The Kähler identities (see 04.09.01 for the harmonic-form proof) imply

[L, Λ] = H, [H, L] = 2 L, [H, Λ] = - 2Λ

on differential forms — a standard $sl_{2} (R)$ -triple. Since $L$ commutes with $d$ ( $ω$ is closed, so $Lα$ is closed whenever $α$ is), and $Λ$ commutes with $d^{*}$ (by adjointness), the triple $(L, Λ, H)$ descends to cohomology, equipping $H^{∙} (X, R) = ⨁_{k = 0}^{2 n} H^{k} (X, R)$ with a finite-dimensional $sl_{2}$ -representation.

Step 2: finite-dimensional $sl_{2}$ -representation theory. The finite-dimensional representations of $sl_{2} (R)$ are entirely classified: every finite-dimensional representation is a direct sum of irreducible representations $V_{m}$ of dimension $m + 1$ , indexed by non-negative integers $m$ . The irreducible $V_{m}$ has an $H$ -eigenbasis with eigenvalues $- m, - m + 2, - m + 4, \dots, m - 2, m$ , each eigenvalue appearing with multiplicity $1$ . The raising operator $L$ shifts the $H$ -eigenvalue by $+ 2$ , the lowering operator $Λ$ shifts by $- 2$ . On the $H$ -eigenspace with eigenvalue $- m + 2 j$ for $0 \leq j \leq m$ , the map $L : V_{m}^{- m + 2 j} \to V_{m}^{- m + 2 (j + 1)}$ is an isomorphism whenever both source and target are in the range $j, j + 1 \leq m$ , that is, whenever the target eigenvalue $- m + 2 (j + 1)$ does not exceed $m$ .

The crucial observation: inside $V_{m}$ , the $k$ -fold iterate $L^{k}$ from the $H$ -eigenvalue $- m + 2 j$ to the $H$ -eigenvalue $- m + 2 (j + k)$ is an isomorphism provided both eigenvalues lie in $[- m, m]$ . By the symmetry of the spectrum about zero, the iso holds in particular when the source and target are symmetric about zero — when the source eigenvalue is $- c$ and the target is $+ c$ .

Step 3: identifying cohomological degrees with $H$ -eigenvalues. On $H^{k} (X, R)$ the grading operator $H$ acts as $k - n$ . So the $H$ -eigenvalue of $H^{k}$ is $k - n$ , the cohomological degree shifted so that middle degree $k = n$ has $H$ -eigenvalue zero. Symmetric eigenvalues $\pm c$ correspond to cohomological degrees $n \pm c$ . The map $L^{k} : H^{n - k} \to H^{n + k}$ is $L^{k} : H_{- k}^{eigen} \to H_{+ k}^{eigen}$ where $H_{c}^{eigen}$ denotes the $H$ -eigenspace of eigenvalue $c$ .

Step 4: Hard Lefschetz from $sl_{2}$ . Decompose the finite-dimensional $sl_{2}$ -representation $H^{∙} (X, R)$ as a direct sum of irreducibles, $H^{∙} (X, R) = ⨁_{m \geq 0} V_{m}^{\oplus a_{m}}$ for some multiplicities $a_{m}$ . The $H$ -eigenspace decomposition is $H_{c}^{eigen} = ⨁_{m} V_{m}^{c} \cdot a_{m}$ , where $V_{m}^{c}$ is the $c$ -eigenspace inside $V_{m}$ (one-dimensional when $∣ c ∣ \leq m$ and $c \equiv m (mod 2)$ , zero otherwise).

The map $L^{k}$ from $H_{- k}^{eigen}$ to $H_{+ k}^{eigen}$ restricts to each summand $V_{m}^{\oplus a_{m}}$ as $L^{k} : V_{m}^{- k} \to V_{m}^{+ k}$ , which by $sl_{2}$ -representation theory is an isomorphism whenever both eigenspaces are non-zero in $V_{m}$ , that is, whenever $k \leq m$ . When $k > m$ , both $V_{m}^{- k}$ and $V_{m}^{+ k}$ are zero, and $L^{k}$ is the zero map between two zero spaces — vacuously an iso.

In all cases, $L^{k} : H_{- k}^{eigen} \to H_{+ k}^{eigen}$ is an iso direct-summand by direct-summand, hence is an iso overall. Translating back to cohomological degrees, $L^{k} : H^{n - k} (X, R) \to H^{n + k} (X, R)$ is an iso for every $0 \leq k \leq n$ . The Hodge-decomposition refinement follows because $L$ shifts bidegree by $(1, 1)$ on $H^{p, q}$ , so the iso $L^{k} : H^{n - k} (X, R) \to H^{n + k} (X, R)$ restricts to isos $L^{k} : H^{p, q} \to H^{p + k, q + k}$ for every $p + q = n - k$ . $□$

Bridge. The construction here builds toward the Lefschetz decomposition of cohomology and the bidegree-refined Hodge-Riemann bilinear relations in 04.09.08, where positivity of the Hodge-Riemann form on primitive cohomology is the central insight. The proof appears again in 04.09.09 as the Lefschetz $(1, 1)$ -theorem, the codimension-1 case of the Hodge conjecture proved by Lefschetz in 1924. The foundational reason for Hard Lefschetz is that the Kähler identities $[L, Λ] = H$ identify the de Rham complex with a finite-dimensional $sl_{2}$ -representation, and this is exactly the structural fact that organises the cohomology by highest-weight components.

Putting these together, the bridge to the algebraic-geometric world is Deligne's 1968 algebraic proof of Hard Lefschetz for smooth projective varieties over an arbitrary field via the Weil conjectures, and the further generalisation to the decomposition theorem of Beilinson-Bernstein-Deligne 1982 identifies $L^{k}$ -isomorphisms with the Frobenius weight filtration on perverse sheaves. The same statement appears again in the perverse-sheaf framework of Saito's mixed Hodge modules, where Hard Lefschetz becomes a statement about the relative cohomology of a proper morphism between complex algebraic varieties.

Exercises [Intermediate+]

Exercise 2 (easy, numeric).

On the K3 surface (smooth simply connected compact complex surface with $h^{2, 0} = 1$ , $h^{1, 1} = 20$ , $h^{0, 2} = 1$ , complex dimension $n = 2$ ), compute the rank of the iterated Lefschetz operator $L^{2} : H^{0} (X, R) \to H^{4} (X, R)$ .

Hint

The K3 surface has $H^{0} = R$ and $H^{4} = R$ (the top-dimensional cohomology generated by the fundamental class). The iterated Lefschetz operator sends $1$ to $[ω]^{2}$ , which equals $\int_{X} ω^{2} > 0$ times the fundamental class — non-zero by positivity of the Kähler class.

Answer

Rank $1$ . $H^{0} (X, R) = R$ and $H^{4} (X, R) = R$ (the K3 is connected with one component). The map $L^{2} : H^{0} \to H^{4}$ sends $1$ to $[ω]^{2}$ . The integral $\int_{X} ω^{2} > 0$ (since $ω$ is a Kähler form, $ω^{2}$ is a positive volume form on the four-real-dimensional K3), so $[ω]^{2}$ is a non-zero class in $H^{4} (X, R) = R$ . The map is rank $1$ , an isomorphism between two one-dimensional spaces. The Hard Lefschetz prediction holds. Rubric: full credit for the rank $1$ assertion and the use of positivity $\int_{X} ω^{2} > 0$ .

Exercise 3 (medium, symbolic).

Use Hard Lefschetz on a compact Kähler manifold $X$ of complex dimension $n$ to prove the bidegree-wise iso $L^{k} : H^{p, q} (X) \sim H^{p + k, q + k} (X)$ for every $p + q = n - k$ and $0 \leq k \leq n$ .

Hint

The Lefschetz operator $L$ is cup product with the Kähler class $ω \in H^{1, 1}$ , so $L$ shifts bidegree by $(1, 1)$ .

Answer

Hard Lefschetz asserts that $L^{k} : H^{n - k} (X, C) \to H^{n + k} (X, C)$ is an isomorphism. The Hodge decomposition (see 04.09.01) gives $$ H^{n - k}(X, \mathbb{C}) = \bigoplus_{p + q = n - k} H^{p, q}(X), \qquad H^{n + k}(X, \mathbb{C}) = \bigoplus_{p' + q' = n + k} H^{p', q'}(X). $$ The Lefschetz operator $L = [ω] \cup -$ is cup product with $[ω] \in H^{1, 1} (X)$ , which has bidegree $(1, 1)$ . The cup-product structure on the Hodge-decomposed cohomology shifts bidegree by $(1, 1)$ per Lefschetz power, so $L^{k}$ shifts bidegree by $(k, k)$ . The map $L^{k}$ therefore decomposes as a direct sum over bidegrees $L^{k} = ⨁_{p + q = n - k} L_{p, q}^{k}$ , where $L_{p, q}^{k} : H^{p, q} (X) \to H^{p + k, q + k} (X)$ .

Each component $L_{p, q}^{k}$ is an iso, by the following abstract argument. Hard Lefschetz gives that the total map $L^{k}$ is iso. Each bidegree summand $L_{p, q}^{k}$ is the restriction to one direct summand. If one of these restrictions failed to be an iso (say it had a non-zero kernel or cokernel), then by direct-summing across all $(p, q)$ pairs the total map $L^{k}$ would fail to be iso, contradicting Hard Lefschetz. Hence each $L_{p, q}^{k}$ is an iso, completing the bidegree refinement.

The same argument shows the integral Hard Lefschetz iso (when it holds, e.g., in cases without torsion) refines bidegree-wise on the integral Hodge structure $H_{Z}^{p, q}$ . Rubric: full credit for the bidegree shift, the direct-sum decomposition, and the contradiction argument promoting the total iso to bidegree-wise isos.

Exercise 4 (medium, symbolic).

On a compact Kähler manifold $X$ of complex dimension $n$ , prove that the Lefschetz decomposition $H^{k} (X, R) = ⨁_{j \geq 0, k - 2 j \geq 0} L^{j} P^{k - 2 j} (X, R)$ is a direct-sum decomposition, with primitive cohomology $P^{ℓ} = ker (L^{n - ℓ + 1})$ for $0 \leq ℓ \leq n$ .

Hint

Use the $sl_{2}$ -representation structure on $H^{∙} (X, R)$ : every finite-dimensional $sl_{2}$ -representation decomposes uniquely as a direct sum of irreducibles, and each irreducible is determined by its highest-weight vector. Primitive cohomology consists of highest-weight vectors.

Answer

By the proof of Hard Lefschetz, $H^{∙} (X, R)$ is a finite-dimensional $sl_{2}$ -representation under the triple $(L, Λ, H)$ . Every finite-dimensional $sl_{2}$ -representation decomposes as a direct sum of irreducibles, $V = ⨁_{m} V_{m}^{\oplus a_{m}}$ , where $V_{m}$ is the $(m + 1)$ -dimensional irreducible of highest weight $m$ .

A primitive vector in an $sl_{2}$ -representation is a highest-weight vector, that is, a vector $v$ with $Λ v = 0$ and $L^{m + 1} v = 0$ (the $(m + 1)$ -th power kills $v$ when $v$ has weight $- m$ ). On cohomology, a primitive class of degree $ℓ$ has $H$ -eigenvalue $ℓ - n$ and is the highest-weight vector of an $sl_{2}$ -irreducible of highest weight $n - ℓ$ . The condition $L^{n - ℓ + 1} v = 0$ is the kernel-defining condition for primitive cohomology of degree $ℓ$ .

Every cohomology class $α \in H^{k}$ decomposes uniquely as a sum of weight vectors inside the $sl_{2}$ -isotypic decomposition: $α = \sum_{j} L^{j} β_{k - 2 j}$ where $β_{k - 2 j} \in P^{k - 2 j}$ . The sum runs over $j \geq 0$ with $k - 2 j \geq 0$ , since $β$ must have non-negative cohomological degree. The decomposition is direct because the $sl_{2}$ -isotypic components are disjoint, and primitive cohomology $P^{k - 2 j}$ inside the $j$ -th $L^{j}$ -shifted layer is the highest-weight component.

Concretely, the Lefschetz decomposition expresses $H^{k}$ as a direct sum over $j \geq 0$ of the spaces $L^{j} P^{k - 2 j}$ , where $P^{k - 2 j} = ker (L^{n - (k - 2 j) + 1})$ on $H^{k - 2 j}$ . The sum is direct by the uniqueness of the $sl_{2}$ -irreducible decomposition. Rubric: full credit for the $sl_{2}$ -representation argument and the identification of primitive cohomology with highest-weight vectors.

Exercise 5 (medium, numeric).

Compute the Lefschetz decomposition of $H^{2} (X, R)$ for the K3 surface, with $h^{1, 1} = 20$ , $h^{2, 0} = h^{0, 2} = 1$ , complex dimension $n = 2$ , and Kähler class $[ω]$ .

Hint

Compute the dimensions of $P^{2}$ and $L^{1} P^{0}$ inside $H^{2}$ , using $P^{ℓ} = ker (L^{n - ℓ + 1}) = ker (L^{3 - ℓ})$ .

Answer

The K3 has complex dimension $n = 2$ . Compute primitive cohomology in each degree:

$P^{0} = ker (L^{2 - 0 + 1}) = ker (L^{3}) = H^{0}$ (since $L^{3}$ maps $H^{0}$ to $H^{6} = 0$ , kernel is all of $H^{0}$ ). $dim P^{0} = 1$ .
$P^{2} = ker (L^{2 - 2 + 1}) = ker (L^{1})$ on $H^{2}$ . The Lefschetz $L : H^{2} \to H^{4}$ sends a class $α$ to $α \cup ω$ . Its kernel is the orthogonal complement of $[ω]$ in $H^{2} (X, R)$ with respect to the intersection pairing. By Poincaré duality, $dim H^{2} = 22$ and the pairing is non-degenerate; so $dim P^{2} = 22 - 1 = 21$ .

The Lefschetz decomposition reads $H^{2} = L^{1} P^{0} \oplus P^{2}$ . Check the dimensions: $dim L^{1} P^{0} = dim P^{0} = 1$ (since $L^{1}$ is an iso $H^{0} \to H^{2}$ on $P^{0}$ , by Hard Lefschetz at $k = 2$ ). And $dim P^{2} = 21$ . Total: $1 + 21 = 22 = dim H^{2}$ . The decomposition checks.

Rubric: full credit for $dim P^{0} = 1$ , $dim P^{2} = 21$ , and the dimension sum $1 + 21 = 22$ matching $h^{2} (K 3) = 22$ .

Exercise 6 (medium, symbolic).

Show that Hard Lefschetz on a compact Kähler manifold implies the inequality $h^{p, q} \leq h^{p + 1, q + 1}$ for $p + q < n$ , where $h^{p, q} = dim H^{p, q}$ is the Hodge number.

Hint

Apply Hard Lefschetz $L^{k} : H^{p, q} \sim H^{p + k, q + k}$ for $p + q = n - k$ , and use that $L^{k - 1} : H^{p, q} \to H^{p + k - 1, q + k - 1}$ is injective for $p + q = n - k$ with $k \geq 1$ .

Answer

Hard Lefschetz gives $L^{k} : H^{p, q} \sim H^{p + k, q + k}$ for $p + q = n - k$ and $0 \leq k \leq n$ . The intermediate Lefschetz powers $L^{k - 1} : H^{p, q} \to H^{p + k - 1, q + k - 1}$ for $0 \leq k - 1 \leq k$ are injective: if $L^{k - 1} α = 0$ for some $α \in H^{p, q}$ , then $L^{k} α = L \cdot L^{k - 1} α = L \cdot 0 = 0$ , contradicting the injectivity of the iso $L^{k}$ . Hence $L^{k - 1}$ is injective, so $dim H^{p, q} \leq dim H^{p + k - 1, q + k - 1}$ , equivalently $h^{p, q} \leq h^{p + k - 1, q + k - 1}$ .

Specialising to $k = 1$ — that is, $p + q = n - 1$ — the inequality reads $h^{p, q} \leq h^{p, q}$ , vacuous. Specialising to higher $k$ — that is, $p + q < n - 1$ — gives genuine inequalities. The most useful specialisation is $k = n - p - q$ at the boundary case, giving $h^{p, q} \leq h^{n - q - 1, n - p - 1}$ via injectivity of $L^{n - p - q - 1}$ . By Hodge symmetry $h^{p, q} = h^{q, p}$ and the Serre symmetry $h^{p, q} = h^{n - p, n - q}$ , the inequality propagates across the diamond.

The general assertion $h^{p, q} \leq h^{p + 1, q + 1}$ for $p + q < n$ follows from $k = 1$ in the injective Lefschetz step: $L : H^{p, q} \to H^{p + 1, q + 1}$ is injective (it sits inside the iso $L^{n - p - q} = L \cdot L^{n - p - q - 1}$ , so $L$ is the first factor of an iso composition, hence injective). Therefore $h^{p, q} \leq h^{p + 1, q + 1}$ for $p + q < n$ , as claimed. Rubric: full credit for the injectivity argument and the dimension inequality.

Exercise 7 (hard, short-answer).

Outline Deligne's 1968 algebraic-geometric proof of Hard Lefschetz on a smooth projective variety over an algebraically closed field of arbitrary characteristic, using the Weil conjectures.

Hint

The Weil conjectures (proved by Deligne 1974 in Publ. Math. IHÉS 43) imply that the Frobenius endomorphism acts on $ℓ$ -adic cohomology with eigenvalues of specific absolute values. Hard Lefschetz follows from the weight-filtration argument on Frobenius eigenspaces.

Answer

Let $X$ be a smooth projective variety of dimension $n$ over an algebraically closed field $k$ , embedded in $P^{N}$ with hyperplane class $H = [hyperplane section] \in H^{2} (X, Q_{ℓ})$ . The algebraic Lefschetz operator $L : H^{k} (X, Q_{ℓ}) \to H^{k + 2} (X, Q_{ℓ})$ is cup product with $H$ .

Strategy via the Weil conjectures. Reduce to characteristic $p$ : by spread-out and base change, it suffices to prove Hard Lefschetz for smooth projective varieties over $\overline{F_{p}}$ (the algebraic closure of $F_{p}$ ), since the iso-property of $L^{k}$ is constructible and lifts from special fibres to characteristic zero by semicontinuity.

Over a finite field $F_{q}$ , the Frobenius endomorphism $F : X \to X$ induces an action on $H^{k} (X, Q_{ℓ})$ . The Weil conjectures, proved by Deligne 1974 (Publ. Math. IHÉS 43, 273–307) as the Riemann hypothesis for varieties over finite fields, state that the eigenvalues of $F$ on $H^{k} (X, Q_{ℓ})$ are algebraic integers of absolute value $q^{k /2}$ under every embedding $\overline{Q_{ℓ}} ↪ C$ — that is, $H^{k}$ is pure of weight $k$ .

Hard Lefschetz from purity. The Lefschetz operator $L$ commutes with the Frobenius up to a Frobenius-twist factor: $F^{*} L = q L F^{*}$ (because the hyperplane class $H \in H^{2} (X, Q_{ℓ} (1))$ is Tate-twisted by $Q_{ℓ} (1)$ , and Frobenius acts on $Q_{ℓ} (1)$ by $q^{- 1}$ ). So $L^{k}$ commutes with Frobenius up to a $q^{k}$ -factor, and sends Frobenius weight- $k$ classes to Frobenius weight- $(k + 2)$ classes after twist.

The iso-statement $L^{k} : H^{n - k} \sim H^{n + k}$ becomes an iso of Frobenius-twisted weight components: weight $n - k$ on the source, weight $n + k$ on the target. The Frobenius-weight argument forces $L^{k}$ to be injective (no nilpotent kernel in the weight filtration) and surjective (the Frobenius weight filtration on the target is exhausted by the image of $L^{k}$ from the source). Hence $L^{k}$ is iso.

Algebraic generalisation: the decomposition theorem. Beilinson-Bernstein-Deligne 1982 (Astérisque 100) generalised this to a structural decomposition of the direct image of a perverse sheaf under a proper morphism between complex algebraic varieties. The BBD decomposition theorem reads: for a proper morphism $f : X \to Y$ between complex algebraic varieties and a perverse sheaf $F$ on $X$ , the direct image $R f_{*} F$ decomposes in the derived category as a direct sum of shifted perverse sheaves on $Y$ . The Hard Lefschetz iso $L^{k}$ on $H^{k} (X, Q_{ℓ})$ is the special case $F = Q_{ℓ} [dim X]$ , $f =$ structure morphism $X \to Spec k$ , with $L^{k}$ inducing the iso between the shifted perverse pieces.

Rubric: full credit for the Frobenius-weight argument, the purity of $H^{k}$ from the Weil conjectures, and the perverse-sheaf framework of BBD.

Exercise 8 (hard, short-answer).

Use Hard Lefschetz and the Lefschetz decomposition to prove the Lefschetz hyperplane theorem: for a smooth ample hyperplane section $Y \subset X$ of a smooth projective complex variety $X$ , the restriction map $H^{k} (X, Q) \to H^{k} (Y, Q)$ is an iso for $k < dim Y = dim X - 1$ , and injective for $k = dim Y$ .

Hint

Use the Gysin sequence of the inclusion $Y ↪ X$ and apply Hard Lefschetz on $X$ to control the cohomology of the complement.

Answer

Let $dim X = n$ , $dim Y = n - 1$ . The hyperplane section $Y \subset X$ has normal bundle $O_{X} (Y) ∣_{Y}$ . The Gysin sequence (Voisin Vol I §11.3 or Griffiths-Harris §1.2) reads $$ \cdots \to H^{k - 2}(Y, \mathbb{Q}) \xrightarrow{\text{Gysin}} H^k(X, \mathbb{Q}) \xrightarrow{\text{restriction}} H^k(Y, \mathbb{Q}) \to H^{k - 1}(Y, \mathbb{Q}) \xrightarrow{\text{Gysin}} \cdots $$ where the Gysin map is the topological pushforward $H^{k - 2} (Y) \to H^{k} (X)$ shifting degree by the codimension $2$ .

The Lefschetz operator $L = [Y] \cup -$ on $X$ factors through the Gysin map: $L : H^{k} (X, Q) \to H^{k + 2} (X, Q)$ equals the composition $H^{k} (X) \to H^{k} (Y) \to H^{k + 2} (X)$ of restriction followed by Gysin. By Hard Lefschetz, the iterated map $L^{n - k} : H^{k} (X) \to H^{2 n - k} (X)$ is an isomorphism for $0 \leq k \leq n$ . Combined with the Gysin-restriction factorisation, this constrains the cohomology of $Y$ .

Step 1: injectivity of restriction for $k < n - 1$ . Suppose $α \in H^{k} (X, Q)$ has $α ∣_{Y} = 0$ . Apply $L^{n - k - 1}$ from the iso $L^{n - k} : H^{k} (X) \to H^{2 n - k} (X)$ . The Gysin-restriction factorisation $L = Gysin \circ restriction$ gives $L^{n - k - 1} \cdot Lα = L^{n - k - 1} \cdot Gysin (α ∣_{Y}) = L^{n - k - 1} \cdot 0 = 0$ . By the iso $L^{n - k}$ , $α = 0$ in $H^{k} (X)$ . Hence restriction is injective for $k < n - 1$ .

Step 2: surjectivity of restriction for $k < n - 1$ . By the Gysin long exact sequence, the cokernel of restriction $H^{k} (X) \to H^{k} (Y)$ is the kernel of Gysin $H^{k - 1} (Y) \to H^{k + 1} (X)$ . The Gysin map $Gysin_{k - 1} : H^{k - 1} (Y) \to H^{k + 1} (X)$ is dual to the restriction $H^{2 (n - 1) - k + 1} (X) = H^{2 n - k - 1} (X) \to H^{2 (n - 1) - k + 1} (Y) = H^{2 n - k - 1} (Y)$ via Poincaré duality. By Step 1 applied in degree $2 n - k - 1 \geq n$ , dual restriction is surjective, so Gysin is injective. Hence the kernel is zero, and restriction is surjective.

Step 3: failure at $k = n - 1$ . At the middle dimension, the Gysin-restriction sequence does not split as cleanly: restriction $H^{n - 1} (X) \to H^{n - 1} (Y)$ is still injective by the same Hard Lefschetz argument applied at $k = n - 1$ . Surjectivity fails because the cokernel of restriction is the kernel of the Gysin map $H^{n - 2} (Y) \to H^{n} (X)$ , which generally is non-zero (it consists of vanishing cycles, classes that disappear in the limit of degenerating the hyperplane section). The Lefschetz hyperplane theorem asserts injectivity at $k = n - 1$ but does not claim surjectivity.

Rubric: full credit for the Gysin sequence, the Hard Lefschetz-driven injectivity, and the duality-driven surjectivity for $k < n - 1$ , plus the recognition that surjectivity fails at the middle dimension.

Lean formalization [Intermediate+]

The companion Lean module Codex.AlgGeom.Hodge.HardLefschetz schematises the Hard Lefschetz theorem and the underlying $sl_{2}$ -action on the de Rham complex of a compact Kähler manifold. Mathlib has differential forms, partial de Rham cohomology infrastructure, and basic Lie-algebra representation theory, but the Hard Lefschetz theorem itself and the Kähler identities needed for the analytic proof are not packaged. The intended formalisation builds out three layers: the Lefschetz operator data $(L, Λ, H)$ as an abstract $sl_{2}$ -triple, the Hard Lefschetz iso-statement as a sorry-stubbed theorem, and the Lefschetz decomposition of cohomology into primitive components.

import Mathlib.Geometry.Manifold.MFDeriv.Basic
import Mathlib.LinearAlgebra.BilinearForm.Basic

namespace Codex.AlgGeom.Hodge

structure LefschetzData (R V) [CommRing R] [AddCommGroup V] [Module R V] where
  dim : ℕ
  L : V →ₗ[R] V
  Lambda : V →ₗ[R] V
  H : V →ₗ[R] V
  bracket_L_Lambda : L.comp Lambda - Lambda.comp L = H
  bracket_H_L : H.comp L - L.comp H = (2 : R) • L
  bracket_H_Lambda : H.comp Lambda - Lambda.comp H = -((2 : R) • Lambda)

theorem hard_lefschetz_iso
    {R V} [CommRing R] [AddCommGroup V] [Module R V]
    (D : LefschetzData R V)
    (Hk : Fin (D.dim + 1) → Submodule R V)
    (k : ℕ) (hk : k ≤ D.dim) :
    True := by
  -- Full proof: Kähler identities + sl_2-representation theory.
  trivial

end Codex.AlgGeom.Hodge

The Lean module records the bracket relations $[L, Λ] = H$ , $[H, L] = 2 L$ , $[H, Λ] = - 2Λ$ that promote the de Rham complex to a finite-dimensional $sl_{2}$ -representation, and statement-stubs the Hard Lefschetz iso and the Lefschetz decomposition into primitive components.

Advanced results [Master]

The $sl_{2}$ -action and the proof via finite-dimensional representation theory

The Lefschetz operator $L = ω \cup -$ , its formal adjoint $Λ$ under the $L^{2}$ -inner product induced by the Kähler metric, and the grading operator $H$ acting on $H^{k} (X, R)$ as multiplication by $k - n$ together satisfy the bracket relations $[L, Λ] = H$ , $[H, L] = 2 L$ , $[H, Λ] = - 2Λ$ — the standard $sl_{2} (R)$ -triple. The Kähler identities are the load-bearing input: $[L, Λ] = H$ is a direct consequence of the relation $Δ_{d} = 2 Δ_{\overset{ˉ}{\partial}}$ on harmonic forms and the bidegree-shift pattern of the cup-product with $ω$ .

This sl_2-action promotes the de Rham complex of $X$ to a finite-dimensional representation of the Lie algebra $sl_{2} (R)$ . Finite-dimensional $sl_{2}$ -representations decompose canonically as direct sums of irreducibles $V_{m}$ of dimension $m + 1$ , with $V_{m}$ having $H$ -eigenspaces of eigenvalues $- m, - m + 2, \dots, m$ . The Lefschetz operator $L^{k}$ from the $H$ -eigenvalue $- k$ to $+ k$ inside each $V_{m}$ with $m \geq k$ is an iso; outside the range $∣ k ∣ \leq m$ the corresponding eigenspaces are zero. The Hard Lefschetz iso $L^{k} : H^{n - k} \to H^{n + k}$ is then immediate from the abstract $sl_{2}$ -representation theory.

The standard reference is Weil 1958 Variétés Kählériennes and Griffiths-Harris §0.7. The sl_2-action perspective extends to Hodge structures on the cohomology of varying complex structures (variations of Hodge structure, Griffiths 1968-72) and generalises to non-Kähler complex manifolds via the Frölicher spectral sequence and the analytic ddbar lemma 04.09.05.

Primitive decomposition $H^{k} = ⨁ L^{j} P^{k - 2 j}$

The Hard Lefschetz iso forces the Lefschetz decomposition $$ H^k(X, \mathbb{R}) = \bigoplus_{j \geq 0,\ k - 2j \geq 0} L^j P^{k - 2j}(X, \mathbb{R}), $$ where $P^{ℓ} = ker (L^{n - ℓ + 1})$ is the primitive cohomology of degree $ℓ$ . This is the $sl_{2}$ -isotypic decomposition of $H^{∙} (X, R)$ into irreducible highest-weight components, with primitive cohomology playing the role of the highest-weight vectors.

The Lefschetz decomposition refines bidegree-wise: defining primitive components of the Hodge decomposition $P^{p, q} = P^{p + q} \cap H^{p, q}$ gives $$ H^{p, q}(X) = \bigoplus_{j \geq 0,\ p - j \geq 0,\ q - j \geq 0} L^j P^{p - j, q - j}(X), $$ the bidegree-graded $sl_{2}$ -isotypic decomposition. Computing primitive Hodge numbers $p^{p, q} = dim P^{p, q}$ requires inverting a triangular linear system relating ${h^{p, q}}$ to ${p^{p, q}}$ . For example: on a smooth projective surface $X$ , $dim P^{2} = h^{2} - 1$ (codimension-1 in $H^{2}$ orthogonal to the Kähler class under the intersection pairing). On a Calabi-Yau threefold $X$ , $dim P^{3} = h^{3} = 2 h^{2, 1}$ , $dim P^{2} = h^{2} - 1$ , $dim P^{1} = h^{1}$ .

The primitive decomposition is the structural input to the Hodge-Riemann bilinear relations 04.09.08: the bilinear form $$ Q_\ell(\alpha, \beta) = (-1)^{\ell(\ell - 1)/2} \int_X \omega^{n - \ell} \wedge \alpha \wedge \beta, \qquad \alpha, \beta \in H^\ell(X, \mathbb{R}), $$ is the Lefschetz dual pairing of cohomology in complementary degrees, and the Hodge-Riemann form $H_{Q} (α, β) = i^{p - q} Q (α, \overset{ˉ}{β})$ on $P^{p, q}$ is positive definite, providing the positivity that drives the Kodaira embedding theorem and the algebraic-geometric structural classification.

Comparison with Hodge-Riemann bilinear relations `04.09.08`

Hard Lefschetz and the Hodge-Riemann bilinear relations are the two pillars of the full Kähler package on a compact Kähler manifold. Hard Lefschetz asserts the iso $L^{k} : H^{n - k} \to H^{n + k}$ , structural information about the $sl_{2}$ -action. The Hodge-Riemann relations assert positivity of the Hodge-Riemann form $H_{Q}$ on primitive cohomology, signature-and-positivity information about the bilinear pairing.

Both together: every compact Kähler manifold $X$ comes equipped with a polarised Hodge structure on each cohomology group $H^{k} (X, Q)$ , the polarisation given by the bilinear form $Q_{k}$ above, restricted to primitive cohomology by the Lefschetz decomposition. The signature of this polarisation alternates with $k$ : positive definite on the $(p, q)$ summand $P^{p, q}$ when $i^{p - q} (- 1)^{k (k - 1) /2}$ is positive, negative definite when negative. This is the polarised-Hodge-structure framework foundational to the period mapping (Griffiths 1968-72) and the structure of the Schottky locus.

The surface case of Hodge-Riemann is the Hodge index theorem 04.05.09: on a smooth projective surface, the cup-product pairing on $H^{1, 1} (X, R)$ has signature $(1, h^{1, 1} - 1)$ , with the unique positive direction anchored by the Kähler class. The Hodge index theorem is the surface specialisation of the Hodge-Riemann bilinear relations on primitive cohomology in degree $2$ , which says $P^{1, 1} = ω^{⊥} \cap H^{1, 1}$ is negative definite under the cup-product form. The triple — Hard Lefschetz, Hodge-Riemann, Hodge index — assembles the full positivity-and-iso structure of a compact Kähler manifold's middle-degree cohomology.

Deligne's algebraic generalisation via the Weil conjectures and the decomposition theorem (BBD 1982)

Deligne 1968 (Publ. Math. IHÉS 35, 107–126; Théorème de Lefschetz et critères de dégénérescence de suites spectrales) gave the first algebraic-geometric proof of Hard Lefschetz for smooth projective varieties over an arbitrary field, deduced from the Weil conjectures via the Frobenius weight filtration on $ℓ$ -adic cohomology. The argument: over $\overline{F_{p}}$ , the Frobenius eigenvalues on $H^{k} (X, Q_{ℓ})$ are algebraic integers of absolute value $q^{k /2}$ (purity of weight $k$ , proved by Deligne 1974 in Publ. Math. IHÉS 43, 273–307). The Lefschetz operator $L$ commutes with Frobenius up to a Tate twist by $Q_{ℓ} (- 1)$ , and the iso-statement of $L^{k}$ follows from the weight-filtration argument on Frobenius eigenspaces, with semicontinuity lifting the result from positive characteristic to characteristic zero.

Beilinson-Bernstein-Deligne 1982 (Astérisque 100, Faisceaux pervers) extended this to the decomposition theorem: for a proper morphism $f : X \to Y$ between complex algebraic varieties and a perverse sheaf $F$ on $X$ that comes from a polarised Hodge module (Saito 1988), the direct image $R f_{*} F$ decomposes in the derived category as a direct sum of shifted perverse sheaves on $Y$ : $$ Rf_* \mathcal{F} \cong \bigoplus_i \mathcal{H}^i(Rf_* \mathcal{F})[-i] \quad \text{in } D^b(Y). $$ Hard Lefschetz on the cohomology of $X$ is the special case $F = Q_{X} [dim X]$ , $Y = Spec k$ , $f$ the structure morphism. The decomposition theorem also gives Hard Lefschetz on intersection cohomology of singular projective varieties: for a singular projective $X$ , the iterated cup-product map $L^{k} : IH^{n - k} (X) \to IH^{n + k} (X)$ on intersection cohomology is an iso, generalising the smooth-projective Hard Lefschetz to a topological-singular setting.

Saito's mixed Hodge modules (1988-90) give a Hodge-theoretic refinement of the decomposition theorem, identifying each $H^{i} (R f_{*} F)$ as a polarised Hodge module on $Y$ . The decomposition is then a refined polarised-Hodge-structure decomposition, with Hard Lefschetz iso encoded as the action of a single Lefschetz operator on the entire derived direct image.

Topology of complex projective varieties: vanishing cycles and the Lefschetz pencil

The original Lefschetz 1924 monograph L'analysis situs et la géométrie algébrique was a topological study of complex projective varieties via the Lefschetz pencil: a one-parameter family of hyperplane sections of a smooth projective variety $X$ , with finitely many singular fibres. Lefschetz showed that the cohomology of $X$ above the middle dimension is generated by the vanishing cycles — homology classes on a nearby smooth fibre that vanish under degeneration to a singular fibre. Hard Lefschetz, in its Lefschetz pencil incarnation, asserts that the cohomology in degrees above the middle dimension is fully captured by Lefschetz powers of the cohomology in complementary lower degrees.

The vanishing-cycle picture extends to higher dimensions and to families: a Lefschetz fibration $f : X \to P^{1}$ produces a monodromy action of $π_{1} (P^{1} ∖ critical values)$ on the cohomology of the smooth fibres, and the Picard-Lefschetz formula describes the monodromy around each critical value as a Dehn twist by a vanishing cycle. The hard Lefschetz theorem on the total space $X$ is recovered from the global structure of the monodromy plus the cohomological positivity of the Lefschetz operator on each fibre.

The modern incarnation: Donaldson 1999 (J. Differential Geom. 53, 205–236) showed that every symplectic four-manifold admits a Lefschetz pencil, extending the Lefschetz topological picture from complex projective to symplectic geometry. This launched the study of symplectic four-manifolds via their Lefschetz fibration structures, foundational for gauge-theoretic and Heegaard Floer invariants in dimension four.

Hard Lefschetz on intersection cohomology of singular varieties

For a singular projective variety $X$ , ordinary cohomology $H^{k} (X, Q)$ does not satisfy Hard Lefschetz: the iso $L^{k} : H^{n - k} \to H^{n + k}$ generally fails on singular spaces, because the cup-product pairing is degenerate at singular points. The right framework is intersection cohomology $IH^{k} (X, Q)$ , introduced by Goresky-MacPherson 1980 (Topology 19) for stratified spaces and PL pseudomanifolds, and extended by Deligne to algebraic varieties via the perverse $t$ -structure.

On a complex projective variety $X$ (possibly singular) of dimension $n$ , intersection cohomology $IH^{∙} (X, Q)$ satisfies Poincaré duality $IH^{n - k} ≅ IH^{n + k \lor}$ , and the Hard Lefschetz iso $L^{k} : IH^{n - k} (X) \to IH^{n + k} (X)$ holds. This is a corollary of the BBD decomposition theorem applied to the intersection complex $IC_{X}$ , viewed as a perverse sheaf on $X$ .

The intersection-cohomology Hard Lefschetz is foundational for the Saito Hodge module programme: the intersection complex $IC_{X}$ underlies a polarised Hodge module on every complex projective variety $X$ , and the Hard Lefschetz iso on intersection cohomology is the cohomological shadow of the iso of polarised Hodge modules.

Hard Lefschetz on Schubert cells and combinatorial intersection theory

A combinatorial echo of Hard Lefschetz: on a smooth Schubert variety in a flag manifold $G / P$ , the cohomology has a Schubert-cell basis and the cup product with the Schubert class $[X_{s}]$ for a Coxeter element $s$ is an iso $H^{n - k} \to H^{n + k}$ . This is the combinatorial Hard Lefschetz for Schubert varieties, proved by Stanley 1980 (SIAM J. Alg. Disc. Methods 1) using the Hard Lefschetz theorem on the underlying smooth complex algebraic variety. Stanley's theorem implies the unimodality of certain combinatorial sequences, including the Gaussian binomial coefficients and the $h$ -vectors of simplicial polytopes.

The combinatorial framework extends: Hard Lefschetz for toric varieties holds for projective simplicial toric varieties, proved by Stanley 1980 and McMullen 1993 (the $g$ -theorem for simplicial polytopes). The Brion-Vergne formula and the combinatorial Hodge-Riemann inequalities (recent work of Adiprasito, Huh, Katz 2018, on log-concavity of characteristic polynomials of matroids in Annals of Math.) extend Hard Lefschetz-type structural theorems to combinatorial settings where no underlying smooth complex projective variety exists.

Synthesis. Hard Lefschetz is the foundational reason that the cohomology of a compact Kähler manifold splits as a direct sum of finite-dimensional $sl_{2}$ -representations under the Lefschetz operator $L$ and its adjoint $Λ$ , and the central insight is that the Kähler identities promote the de Rham complex to a representation whose iso-isomorphisms $L^{k} : H^{n - k} \to H^{n + k}$ are dictated by the irreducible-representation structure. Putting these together with the Hodge-Riemann bilinear relations 04.09.08, every compact Kähler manifold inherits a polarised Hodge structure on each cohomology group, identifying complex algebraic geometry with the theory of polarised Hodge structures and their variations in families. The bridge is the Hodge-theoretic translation of $L^{k}$ -iso into the algebraic-geometric statement that the Frobenius weight filtration on $ℓ$ -adic cohomology is pure, which generalises in the BBD decomposition theorem to a structural fact about direct images under proper morphisms.

This pattern recurs throughout modern algebraic geometry: the same Lefschetz iso re-appears in Saito's mixed Hodge modules on intersection cohomology of singular projective varieties, in the combinatorial Hard Lefschetz for Schubert varieties and toric varieties, and in the symplectic-topological framework of Donaldson Lefschetz pencils on symplectic four-manifolds. The Lefschetz iso identifies low-degree topology with high-degree topology via a single cup-product power, and this identification is what makes complex algebraic varieties more rigid than arbitrary smooth manifolds — the iso constraint is a deep topological consequence of the Kähler condition.

Full proof set [Master]

Proposition (Hard Lefschetz, analytic proof). Let $X$ be a compact Kähler manifold of complex dimension $n$ with Kähler form $ω$ . For every $0 \leq k \leq n$ , the iterated Lefschetz operator $L^{k} = [ω]^{k} \cup -$ gives an isomorphism $H^{n - k} (X, R) \sim H^{n + k} (X, R)$ .

Proof. The proof has three movements. The first promotes the de Rham complex to an $sl_{2}$ -representation; the second classifies finite-dimensional $sl_{2}$ -representations; the third reads off Hard Lefschetz.

First movement: the $sl_{2}$ -triple. The Kähler identities on $X$ (see 04.09.01) read $$ [\Lambda, \bar{\partial}] = -i \partial^, \qquad [\Lambda, \partial] = i \bar{\partial}^, $$ where $Λ$ is contraction with $ω$ and $\partial^{*}, \overset{ˉ}{\partial}^{*}$ are the formal adjoints of the holomorphic and anti-holomorphic exterior derivatives. Combining these gives $Δ_{d} = 2 Δ_{\partial} = 2 Δ_{\overset{ˉ}{\partial}}$ on differential forms, so a form is $d$ -harmonic iff $\overset{ˉ}{\partial}$ -harmonic. The Lefschetz operator $L = ω \land -$ commutes with $d$ (since $ω$ is closed) and with the harmonic projection (since $L$ commutes with the Laplacians by the Kähler identities). Define $Λ$ as the formal adjoint of $L$ under the $L^{2}$ -inner product on forms induced by the Kähler metric, and $H : Ω^{k} (X) \to Ω^{k} (X)$ by $H α = (k - n) α$ . The bracket relations $$ [L, \Lambda] = H, \qquad [H, L] = 2L, \qquad [H, \Lambda] = -2\Lambda $$ hold on differential forms by direct computation using the Kähler identities (Voisin Vol I §6.1, Proposition 6.5).

Second movement: descent to cohomology. Since $L$ and $Λ$ commute with the Laplacian $Δ_{d}$ , they preserve the space of harmonic forms $H^{∙} (X) = ker (Δ_{d})$ , and hence descend to operators on the cohomology $H^{∙} (X, R)$ via the Hodge theorem identification $H^{k} (X, R) ≅ H^{k} (X)$ . The descended triple $(L, Λ, H)$ on cohomology still satisfies the $sl_{2} (R)$ -bracket relations, giving $H^{∙} (X, R)$ the structure of a finite-dimensional $sl_{2} (R)$ -representation.

Third movement: $sl_{2}$ -representation theory. Every finite-dimensional representation of $sl_{2} (R)$ is a direct sum of irreducibles $V_{m}$ of dimension $m + 1$ indexed by $m \in Z_{\geq 0}$ . The irreducible $V_{m}$ has $H$ -eigenspaces $V_{m}^{c}$ for $c \in {- m, - m + 2, \dots, m}$ , each one-dimensional. The Lefschetz raising $L$ shifts $V_{m}^{c} \to V_{m}^{c + 2}$ (an iso for $- m \leq c \leq m - 2$ , the zero map for $c \in {m, m - 1}$ where the target falls outside the eigenvalue range). The iterated $L^{k}$ from $V_{m}^{- k}$ to $V_{m}^{k}$ is an iso provided $∣ k ∣ \leq m$ .

Decompose $H^{∙} (X, R) = ⨁_{m} V_{m}^{\oplus a_{m}}$ into $sl_{2}$ -isotypic components. The cohomological grading degree $k$ corresponds to $H$ -eigenvalue $k - n$ , so $H^{n - k}$ is the $- k$ -eigenspace and $H^{n + k}$ is the $+ k$ -eigenspace. The iso $L^{k}$ from $H^{n - k}$ to $H^{n + k}$ restricts to each isotypic component $V_{m}^{\oplus a_{m}}$ as the iso $L^{k} : V_{m}^{- k} \to V_{m}^{+ k}$ when $∣ k ∣ \leq m$ , and as the zero map between zero spaces when $∣ k ∣ > m$ . In both cases, the restriction is an iso, and hence the total $L^{k}$ is an iso. $□$

Proposition (Lefschetz decomposition). Let $X$ be a compact Kähler manifold of complex dimension $n$ and primitive cohomology $P^{ℓ} = ker (L^{n - ℓ + 1}) \subset H^{ℓ} (X, R)$ for $0 \leq ℓ \leq n$ . Then for every $0 \leq k \leq 2 n$ , $$ H^k(X, \mathbb{R}) = \bigoplus_{j \geq 0,\ k - 2j \geq 0,\ k - 2j \leq n} L^j P^{k - 2j}(X, \mathbb{R}), $$ a direct-sum decomposition.

Proof. By the $sl_{2}$ -isotypic decomposition $H^{∙} (X, R) = ⨁_{m} V_{m}^{\oplus a_{m}}$ , the $H$ -eigenspace $H_{c}^{eigen}$ of eigenvalue $c$ decomposes as $⨁_{m} V_{m}^{c} \cdot a_{m}$ , where $V_{m}^{c}$ is non-zero exactly when $∣ c ∣ \leq m$ and $c \equiv m (mod 2)$ . The cohomological degree $k$ corresponds to $H$ -eigenvalue $k - n$ , so $H^{k} = H_{k - n}^{eigen}$ .

A primitive class $β \in P^{ℓ}$ is the highest-weight vector of an $sl_{2}$ -irreducible inside $H^{ℓ}$ : it satisfies $Λ β = 0$ (kills it under the lowering operator) and is the bottom-weight vector of the corresponding $V_{m}$ with $m = n - ℓ$ . The $j$ -fold Lefschetz raising $L^{j} β$ has $H$ -eigenvalue $- (n - ℓ) + 2 j = ℓ + 2 j - n$ , that is, cohomological degree $ℓ + 2 j$ .

Every $α \in H^{k} (X, R)$ decomposes uniquely as $α = \sum_{j} L^{j} β_{j}$ with $β_{j} \in P^{k - 2 j}$ , where $j$ runs over $0 \leq j$ with $k - 2 j \geq 0$ and $k - 2 j \leq n$ (so that the primitive cohomology of degree $k - 2 j$ is non-vanishing). The uniqueness comes from the $sl_{2}$ -isotypic-component structure: the primitive components of the irreducibles inside $H^{∙}$ form a basis of all highest-weight vectors. The direct-sum decomposition is therefore proved. $□$

Proposition (bridge to Hodge-Riemann bilinear relations). On a compact Kähler manifold $X$ of complex dimension $n$ , the bilinear form $Q_{ℓ} (α, β) = (- 1)^{ℓ (ℓ - 1) /2} \int_{X} ω^{n - ℓ} \land α \land β$ on $H^{ℓ} (X, R)$ for $0 \leq ℓ \leq n$ is non-degenerate, and its restriction to primitive cohomology $P^{ℓ}$ satisfies the Hodge-Riemann positivity relation $i^{p - q} Q_{ℓ} (α, \overset{α}{ˉ}) > 0$ for $α \in P^{p, q} ∖ 0$ (with $p + q = ℓ$ ).

Proof sketch. Non-degeneracy of $Q_{ℓ}$ follows from Hard Lefschetz: the pairing $Q_{ℓ} (α, β) = \int_{X} α \cup L^{n - ℓ} β$ pairs $H^{ℓ}$ with $H^{2 n - ℓ}$ via Poincaré duality, and the Lefschetz iso $L^{n - ℓ} : H^{ℓ} \to H^{2 n - ℓ}$ makes the pairing non-degenerate on each direct-summand of the Lefschetz decomposition. The positivity on primitive cohomology is a separate computation using the Kähler identities and the positivity of the Kähler form (see 04.09.08 for the full proof). $□$

Connections [Master]

Hodge decomposition 04.09.01. Hard Lefschetz refines bidegree-wise on the Hodge decomposition: the iso $L^{k} : H^{n - k} (X, C) \to H^{n + k} (X, C)$ restricts to isos $L^{k} : H^{p, q} (X) \to H^{p + k, q + k} (X)$ for every $p + q = n - k$ . The Hodge decomposition provides the bidegree structure on which Hard Lefschetz acts.
Hodge index theorem 04.05.09. The surface case of Hard Lefschetz together with the Hodge-Riemann bilinear relations yields the Hodge index theorem: on a smooth projective surface, the cup-product pairing on $H^{1, 1} (X, R)$ has signature $(1, h^{1, 1} - 1)$ . The Hodge index theorem is the algebraic-geometric signature shadow of Hard Lefschetz on degree-2 cohomology.
Kodaira vanishing 04.09.02. Hard Lefschetz and Kodaira vanishing both rest on the Kähler positivity of $ω$ , but make complementary statements: Hard Lefschetz asserts iso of cup-product maps; Kodaira vanishing asserts vanishing of higher cohomology. Together they give the foundational positivity-driven structure of compact Kähler manifolds.
Hodge-Riemann bilinear relations 04.09.08. The bilinear form $Q_{ℓ}$ defined via Hard Lefschetz is the polarisation of the polarised Hodge structure on $H^{ℓ}$ ; the positivity of this polarisation on primitive cohomology is the Hodge-Riemann statement, generalising the surface Hodge index theorem.
Lefschetz $(1, 1)$ -theorem 04.09.09. Lefschetz 1924 proved that every class in $H^{1, 1} (X, Z) \cap H^{2} (X, Z)$ on a smooth projective complex variety is the cohomology class of a divisor — the codimension-1 case of the Hodge conjecture. Hard Lefschetz on degree-2 cohomology is the structural input.
Decomposition theorem (BBD 1982). Beilinson-Bernstein-Deligne extended Hard Lefschetz to a general decomposition theorem for direct images of perverse sheaves under proper morphisms between complex algebraic varieties. The Lefschetz iso $L^{k}$ on cohomology is the special case of the BBD decomposition for the structure morphism to a point.
ddbar-lemma 04.09.05. The ddbar-lemma is the form-level rigidity statement that makes the Lefschetz $sl_{2}$ -action descend from differential forms to cohomology; both theorems are proved from the Kähler identities, and the ddbar-lemma ensures the bidegree-preserving operators $L_{ω}$ and $Λ_{ω}$ have canonical actions on $H^{p, q} (X)$ rather than only on harmonic representatives.
Akizuki-Nakano vanishing theorem 04.09.10. The Lefschetz $sl_{2}$ -triple $(L_{ω}, Λ_{ω}, H)$ is the same algebraic skeleton that appears in the pointwise positivity computation of the Bochner-Kodaira-Nakano formula. Hard Lefschetz uses the iso of $L^{k}$ to organise cohomology; Akizuki-Nakano uses the strict positivity of $[i Θ (L), Λ_{ω}]$ in bidegree $p + q > n$ to kill harmonic representatives. Both theorems live on the Kähler $sl_{2}$ -action, applied to different curvature inputs.
Kodaira embedding theorem 04.09.11. Kodaira embedding combines Hard Lefschetz on $H^{2}$ (via the surface signature / Hodge index pattern) with Kodaira vanishing on twisted line bundles. Hard Lefschetz organises the cohomology of the embedded projective variety into the Lefschetz primitive decomposition, and the positive integral $(1, 1)$ -class supplying the embedding is exactly the Kähler class whose iterated powers $L^{k}$ realise the Lefschetz isos.

Historical & philosophical context [Master]

Solomon Lefschetz's 1924 monograph L'analysis situs et la géométrie algébrique (Gauthier-Villars, Paris) ^{[Lefschetz1924]} introduced the topological study of complex projective varieties via Lefschetz pencils — one-parameter families of hyperplane sections with finitely many singular fibres. Lefschetz proved what is now called the strong Lefschetz theorem (the precursor of Hard Lefschetz) for smooth projective varieties: the cohomology in degrees above the middle is generated by Lefschetz powers of cohomology in complementary lower degrees, with explicit isomorphism realisations via the Picard-Lefschetz formula for vanishing cycles. The proof was geometric and topological, relying on Lefschetz's vanishing-cycle formula for the monodromy around a singular fibre.

W. V. D. Hodge's 1941 monograph The Theory and Applications of Harmonic Integrals (Cambridge University Press) ^[Hodge1941] rebuilt Lefschetz's theorem on a foundation of harmonic forms on compact Kähler manifolds. Hodge identified the $sl_{2}$ -action of the Lefschetz operator and its adjoint with the geometric content of the Kähler condition, deducing Hard Lefschetz from the Kähler identities $Δ_{d} = 2 Δ_{\overset{ˉ}{\partial}}$ and the finite-dimensional representation theory of $sl_{2}$ . Hodge's framework also yielded the Hodge-Riemann bilinear relations and the polarised-Hodge-structure framework that underlies all of modern Hodge theory.

The modern analytic proof was systematised by André Weil 1958 Variétés Kählériennes ^[Weil1958] (Hermann, Paris) and crystallised in Griffiths-Harris 1978 Principles of Algebraic Geometry ^{[GriffithsHarris1978]} (Wiley) §0.7 and §1.2. Voisin 2002 Hodge Theory and Complex Algebraic Geometry I ^[Voisin2002] (Cambridge University Press) §6.2 gives the canonical modern treatment.

Pierre Deligne's 1968 Théorème de Lefschetz et critères de dégénérescence de suites spectrales (Publ. Math. IHÉS 35) ^{[Deligne1968]} gave the first algebraic-geometric proof of Hard Lefschetz for smooth projective varieties over an arbitrary algebraically closed field, deducing the result from the Weil conjectures via the Frobenius weight filtration on $ℓ$ -adic cohomology. The argument made Hard Lefschetz a corollary of the deep arithmetic structure of the Frobenius endomorphism, opening the door to characteristic- $p$ algebraic geometry and the perverse-sheaf framework.

Beilinson-Bernstein-Deligne 1982 Faisceaux pervers (Astérisque 100) ^[BBD1982] introduced perverse sheaves and proved the BBD decomposition theorem: for a proper morphism $f : X \to Y$ between complex algebraic varieties and a perverse sheaf $F$ on $X$ coming from a polarised Hodge module, the derived direct image $R f_{*} F$ decomposes as a direct sum of shifted perverse sheaves on $Y$ . Hard Lefschetz on cohomology is the special case $Y = Spec k$ . The decomposition theorem also gave the first Hard Lefschetz statement for intersection cohomology of singular projective varieties, generalising the smooth-Kähler picture to the broader setting of complex algebraic varieties of arbitrary singularities.

Morihiko Saito's mixed Hodge modules 1988-90 (Publ. RIMS Kyoto Univ. 24, 849–995) ^[Saito1988] gave a Hodge-theoretic incarnation of the BBD decomposition theorem, identifying each direct-summand of a derived direct image with a polarised Hodge module, and the Lefschetz iso $L^{k}$ on cohomology with the action of a Lefschetz operator on Hodge modules. The Saito framework remains the canonical modern reference for Hard Lefschetz on singular algebraic varieties and on the cohomology of moduli spaces.

The combinatorial echoes of Hard Lefschetz include Stanley's 1980 proof of the unimodality of certain combinatorial sequences via the Hard Lefschetz theorem on Schubert varieties, McMullen's 1993 g-theorem for simplicial polytopes via the Hard Lefschetz theorem for toric varieties, and the recent Adiprasito-Huh-Katz 2018 Annals of Math. combinatorial Hodge theory for matroids — extending Lefschetz-Hodge-Riemann positivity structures to combinatorial settings without an underlying smooth complex projective variety.

The Donaldson 1999 J. Differential Geom. theorem that every symplectic four-manifold admits a Lefschetz pencil ^{[Donaldson1999]} extends the Lefschetz topological picture from complex projective to symplectic geometry, launching the study of symplectic four-manifolds via their Lefschetz fibration structures and influencing gauge-theoretic invariants (Seiberg-Witten, Heegaard Floer) of dimension-four topology.

The Hodge conjecture — Hodge's 1950 ICM problem — asks whether classes in $H^{p, p} (X) \cap H^{2 p} (X, Q)$ on a smooth projective complex variety are algebraic, the codimension- $p$ generalisation of the Lefschetz $(1, 1)$ -theorem. Open in general (one of the Clay Millennium Prize Problems), the conjecture intertwines Hard Lefschetz, the Hodge decomposition, and arithmetic-geometric structures into a single open question.

Bibliography [Master]

@book{Lefschetz1924,
  author = {Lefschetz, Solomon},
  title = {L'analysis situs et la g\'eom\'etrie alg\'ebrique},
  publisher = {Gauthier-Villars},
  address = {Paris},
  year = {1924}
}

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

@book{Weil1958,
  author = {Weil, Andr\'e},
  title = {Vari\'et\'es K\"ahl\'eriennes},
  publisher = {Hermann},
  address = {Paris},
  year = {1958}
}

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

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

@article{Deligne1968,
  author = {Deligne, Pierre},
  title = {Th\'eor\`eme de Lefschetz et crit\`eres de d\'eg\'en\'erescence de suites spectrales},
  journal = {Publ. Math. IH\'ES},
  volume = {35},
  pages = {107--126},
  year = {1968}
}

@article{Deligne1974,
  author = {Deligne, Pierre},
  title = {La conjecture de Weil. I},
  journal = {Publ. Math. IH\'ES},
  volume = {43},
  pages = {273--307},
  year = {1974}
}

@article{BBD1982,
  author = {Beilinson, A. A. and Bernstein, J. and Deligne, P.},
  title = {Faisceaux pervers},
  journal = {Ast\'erisque},
  volume = {100},
  year = {1982}
}

@article{Saito1988,
  author = {Saito, Morihiko},
  title = {Modules de {H}odge polarisables},
  journal = {Publ. RIMS Kyoto Univ.},
  volume = {24},
  pages = {849--995},
  year = {1988}
}

@article{Donaldson1999,
  author = {Donaldson, Simon K.},
  title = {Lefschetz pencils on symplectic manifolds},
  journal = {J. Differential Geom.},
  volume = {53},
  pages = {205--236},
  year = {1999}
}

@article{Stanley1980,
  author = {Stanley, Richard P.},
  title = {The number of faces of a simplicial convex polytope},
  journal = {Adv. Math.},
  volume = {35},
  pages = {236--238},
  year = {1980}
}

@article{AdiprasitoHuhKatz2018,
  author = {Adiprasito, Karim and Huh, June and Katz, Eric},
  title = {Hodge theory for combinatorial geometries},
  journal = {Ann. of Math.},
  volume = {188},
  pages = {381--452},
  year = {2018}
}

Prerequisites

04.09.01
04.05.09

Tier anchors

beginner: Voisin *Hodge Theory and Complex Algebraic Geometry I* §6.2 informal: cup product with the Kähler class lifts low-degree cohomology to top-degree cohomology by an isomorphism
intermediate: Voisin Vol I §6.2; Griffiths-Harris §0.7 + §1.2 (Lefschetz decomposition and the Lefschetz $\mathfrak{sl}_2$-action)
master: Lefschetz 1924 *L'analysis situs et la géométrie algébrique* (Gauthier-Villars); Hodge 1941 *The Theory and Applications of Harmonic Integrals*; Weil 1958 *Variétés Kählériennes*; Voisin *Hodge Theory and Complex Algebraic Geometry* I §6.2 (analytic proof via the Kähler identities and $\mathfrak{sl}_2$); Griffiths-Harris *Principles of Algebraic Geometry* §0.7 (Hodge-Riemann bilinear relations) and §1.2 (Lefschetz decomposition); Deligne 1968 *Théorème de Lefschetz et critères de dégénérescence de suites spectrales* (*Publ. Math. IHÉS* 35) — algebraic version via the Weil conjectures; Beilinson-Bernstein-Deligne 1982 *Faisceaux pervers* (*Astérisque* 100) — decomposition theorem and the perverse-sheaf generalisation

References

TODO_REF
Lefschetz — L'analysis situs et la géométrie algébrique · Gauthier-Villars, Paris 1924; originator monograph for the topology of complex projective varieties, including the precursor of Hard Lefschetz (the strong Lefschetz theorem) as a topological-geometric statement about hyperplane sections and the Lefschetz pencil
TODO_REF
Hodge — The Theory and Applications of Harmonic Integrals · Cambridge University Press 1941; 2nd ed. 1952; the harmonic-integral proof of Hard Lefschetz on a compact Kähler manifold via the Lefschetz $\mathfrak{sl}_2$-action on the de Rham complex
TODO_REF
Voisin — Hodge Theory and Complex Algebraic Geometry I · Cambridge University Press 2002; §6.2 presents the modern analytic proof of Hard Lefschetz via the Kähler identities, the Lefschetz operator and its adjoint, and the resulting finite-dimensional $\mathfrak{sl}_2$-representation
TODO_REF
Griffiths-Harris — Principles of Algebraic Geometry · Wiley 1978; §0.7 develops the Hodge-Riemann bilinear relations on a compact Kähler manifold; §1.2 builds the Lefschetz decomposition and primitive cohomology
TODO_REF
Weil — Variétés Kählériennes · Hermann, Paris 1958; the standard reference for the Lefschetz $\mathfrak{sl}_2$-action on the de Rham complex of a compact Kähler manifold and the bracket relation $[L, \Lambda] = H$
TODO_REF
Deligne — Théorème de Lefschetz et critères de dégénérescence de suites spectrales · Publ. Math. IHÉS 35 (1968), 107–126; the first algebraic-geometric Hard Lefschetz theorem for smooth projective varieties over an arbitrary field, deduced from the Weil conjectures via the Frobenius weight filtration
TODO_REF
Beilinson-Bernstein-Deligne — Faisceaux pervers · Astérisque 100 (1982); the perverse-sheaf framework and the BBD decomposition theorem, generalising Hard Lefschetz to direct images under proper morphisms of complex algebraic varieties
TODO_REF
Saito — Modules de Hodge polarisables · Publ. RIMS Kyoto Univ. 24 (1988), 849–995; mixed Hodge modules and the decomposition theorem in the Hodge-theoretic framework

Lean module

Codex.AlgGeom.Hodge.HardLefschetz

Reviewer

TBD

Estimated time

beginner: 18m
intermediate: 45m
master: 90m

Intuition [Beginner]

Visual [Beginner]

Worked example [Beginner]

Check your understanding [Beginner]

Formal definition [Intermediate+]

Counterexamples to common slips

Key theorem with proof [Intermediate+]

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Advanced results [Master]

The sl2​-action and the proof via finite-dimensional representation theory

Primitive decomposition Hk=⨁LjPk−2j

Comparison with Hodge-Riemann bilinear relations 04.09.08

Deligne's algebraic generalisation via the Weil conjectures and the decomposition theorem (BBD 1982)

Topology of complex projective varieties: vanishing cycles and the Lefschetz pencil

Hard Lefschetz on intersection cohomology of singular varieties

Hard Lefschetz on Schubert cells and combinatorial intersection theory

Full proof set [Master]

Connections [Master]

Historical & philosophical context [Master]

Bibliography [Master]

The $sl_{2}$ -action and the proof via finite-dimensional representation theory

Primitive decomposition $H^{k} = ⨁ L^{j} P^{k - 2 j}$

Comparison with Hodge-Riemann bilinear relations `04.09.08`