03.02.25 · differential-geometry / manifolds

The Lefschetz hyperplane theorem via Morse theory

shipped3 tiersLean: none

Anchor (Master): Milnor Morse Theory §7; Andreotti-Frankel 1959; Bott 1959 On a theorem of Lefschetz; Lazarsfeld Positivity in Algebraic Geometry I §3.1

Intuition Beginner

A curve drawn on a flat page and a slice of that page along a straight line look like very different objects — one is two-dimensional, one is one-dimensional. But in complex geometry a surprising rigidity appears: when you cut a complex shape with a flat "knife" (a hyperplane), the slice you get already knows almost everything about the whole shape, up to a dimension that depends on how big the shape is.

This is the Lefschetz hyperplane theorem. It says that a generic flat slice of a complex projective shape is a faithful low-dimensional shadow: holes of small size in the slice and in the whole object match up exactly. Only at the very top dimension can the whole object have features the slice misses.

The reason, found by Andreotti and Frankel, is a counting fact about hills and valleys. If you remove the slice, what is left is an affine piece, and on it a natural "distance from a point" function has only gentle critical behaviour: at every level spot, the number of downhill directions is at most the complex dimension. That single bound forces the leftover piece to be low-dimensional in shape, which is exactly what makes the slice such a good shadow.

Visual Beginner

Picture a complex curve sitting inside complex projective space, and a hyperplane cutting it in a few points. Removing those points leaves an affine curve. On that affine piece, mark the level spots of the squared-distance-to-a-point function.

Every marked spot has a downhill count of at most $n$ , the complex dimension. Because no level spot ever has more than $n$ downhill directions, the affine piece can be rebuilt using only cells of dimension up to $n$ , never higher. That low-dimensional skeleton is the whole secret: it makes the slice $Y$ a faithful copy of $X$ below dimension $n$ .

Worked example Beginner

Take the simplest case: $X$ is the complex projective plane $CP^{2}$ , which has complex dimension $2$ , and the slice $Y$ is a line $CP^{1}$ inside it. The theorem predicts the inclusion of the line into the plane is $(2 - 1) = 1$ -connected: it should match all holes below dimension $1$ , and at least cover dimension $1$ .

Count the real-dimensional holes. The projective line $CP^{1}$ is a sphere: it has one piece (dimension $0$ ) and one two-dimensional cavity, with nothing in dimension $1$ . The projective plane $CP^{2}$ also has one piece, nothing in dimension $1$ , and one two-dimensional cavity. So in dimensions $0$ and $1$ the two objects agree perfectly — one piece each, no one-dimensional holes — exactly as $1$ -connectedness demands.

What this tells us: even though $CP^{2}$ has extra structure higher up (it has a four-dimensional cavity the line cannot see), everything up to dimension $1$ is already visible in the line. The slice is a faithful shadow of the whole below the cutoff, and the cutoff is governed by the complex dimension.

Check your understanding Beginner

Formal definition Intermediate+

Let $X \subset CP^{N}$ be a smooth complex projective variety of complex dimension $n$ (a compact complex submanifold), and let $H \subset CP^{N}$ be a hyperplane. The hyperplane section is $Y = X \cap H$ . The hyperplane is generic (or transverse to $X$ ) if $H$ meets $X$ transversally, so that $Y$ is again a smooth complex submanifold, this time of complex dimension $n - 1$ ; by the Bertini theorem the set of such $H$ is the complement of a proper algebraic subset of the dual projective space $(CP^{N})^{\lor}$ , hence generic. The affine complement is $U = X ∖ Y = X \cap (CP^{N} ∖ H)$ . Choosing affine coordinates so that $H$ is the hyperplane at infinity realises $U$ as a closed complex submanifold of the affine space $C^{N} ≅ CP^{N} ∖ H$ ; such a closed complex submanifold of $C^{N}$ is a smooth affine variety, and in particular a Stein manifold.

A continuous map $i : Y \to X$ is $k$ -connected if it induces isomorphisms $i_{*} : π_{j} (Y, y) \to π_{j} (X, y)$ for all $j < k$ and base points $y$ , and a surjection $i_{*} : π_{k} (Y, y) \to π_{k} (X, y)$ . Equivalently, the relative homotopy groups $π_{j} (X, Y)$ vanish for $j \leq k$ . By the relative Hurewicz theorem, when $Y ↪ X$ is the inclusion of a closed submanifold with $π_{1}$ controlled, $k$ -connectedness is equivalent to the vanishing $H_{j} (X, Y; Z) = 0$ for $j \leq k$ , and dually to $i_{*} : H_{j} (Y) \to H_{j} (X)$ being an isomorphism for $j < k$ and a surjection for $j = k$ . The reading of the theorem we prove is the homological one, with the homotopy upgrade obtained through Hurewicz.

The Morse-theoretic input is the squared-distance function. For a closed submanifold $U \subset C^{M} ≅ R^{2 M}$ and a point $a \in C^{M}$ , set $$ \rho_a : U \to \mathbb{R}, \qquad \rho_a(z) = |z - a|^2 = \sum_{j=1}^{M} |z_j - a_j|^2. $$ This is a smooth, proper, bounded-below exhausting function on the closed (hence properly embedded) submanifold $U$ : the sublevel sets $U^{c} = {ρ_{a} \leq c}$ are compact. A point $p \in U$ is critical for $ρ_{a}$ iff $a - p$ is normal to $U$ at $p$ . The convention here is the standard Euclidean one with $ρ_{a} \geq 0$ ; sign choices for the Hessian follow accordingly.

Key theorem with proof Intermediate+

The proof factors through one analytic estimate — the Andreotti-Frankel index bound — and then a purely topological deduction. It follows Milnor ^{[Milnor §7]} and the original argument of Andreotti and Frankel ^{[Andreotti-Frankel]}, with the distance-function Morse theory as developed in Bott's lectures ^{[fasttrack-texts Morse theory of the distance function]}.

Theorem (Lefschetz hyperplane theorem, homological form). Let $X \subset CP^{N}$ be a smooth projective variety of complex dimension $n$ and $Y = X \cap H$ a smooth generic hyperplane section. Then the inclusion $i : Y ↪ X$ induces isomorphisms $i_{*} : H_{k} (Y; Z) \to H_{k} (X; Z)$ for $k < n - 1$ and a surjection for $k = n - 1$ ; equivalently $H_{k} (X, Y; Z) = 0$ for $k \leq n - 1$ .

The engine is the following statement about the affine complement.

Proposition (Andreotti-Frankel). A smooth affine variety $U$ of complex dimension $n$ has the homotopy type of a CW complex of real dimension $\leq n$ . Consequently $H_{k} (U; Z) = 0$ for $k > n$ , and $H_{k} (U; Z)$ is free for $k = n$ .

Proof of the Proposition. Embed $U$ as a closed complex submanifold of $C^{M}$ (any affine embedding works). For $a \in C^{M}$ consider $ρ_{a} (z) = ∣ z - a ∣^{2}$ . It is proper and exhausting on the closed submanifold $U$ . For generic $a$ the function $ρ_{a}$ is Morse: writing the endpoint/normal-exponential map $E (p, v) = p + v$ on the normal bundle $N U$ , the focal points of $U$ are the critical values of $E$ , a measure-zero set by Sard's theorem, and $ρ_{a}$ is Morse precisely when $a$ avoids them. Fix such an $a$ . The whole content is the index bound:

Index bound. At every critical point $p$ of $ρ_{a}$ , the Morse index satisfies $λ (p) \leq n$ .

Granting this, the handle-attachment theorem 03.02.31 applies to the exhausting Morse function $ρ_{a}$ : $U$ has the homotopy type of a CW complex with one cell of dimension $λ (p) \leq n$ for each critical point $p$ . A CW complex with cells only in real dimensions $\leq n$ has $H_{k} = 0$ for $k > n$ and free $H_{n}$ (no $(n + 1)$ -cells can create torsion or relations in top degree). This proves the Proposition modulo the index bound.

Proof of the index bound. Work at a critical point $p \in U$ , and let $T_{p} U \subset C^{M}$ be the (real $2 n$ -dimensional) tangent space, which is a complex subspace: $J (T_{p} U) = T_{p} U$ , where $J$ is multiplication by $i$ . The Hessian of $ρ_{a}$ at $p$ on the ambient $C^{M} = R^{2 M}$ is $2 Id$ (since $ρ_{a}$ is a sum of squares), positive definite. Restricting a function to a submanifold modifies the Hessian by the second fundamental form: for $v, w \in T_{p} U$ , $$ \mathrm{Hess}(\rho_a|_U)_p(v,w) = \mathrm{Hess}(\rho_a)_p(v,w) - \langle \nabla \rho_a(p), , \mathrm{II}(v,w) \rangle, $$ where $II : T_{p} U \times T_{p} U \to N_{p} U$ is the second fundamental form (symmetric, $N_{p} U$ -valued) and $\nabla ρ_{a} (p) = 2 (p - a)$ is normal to $U$ at the critical point $p$ . The first term is $2 ⟨ v, w ⟩$ , the ambient positive-definite metric restricted to $T_{p} U$ . Define the symmetric real bilinear form $$ B(v, w) := \langle , p - a, , \mathrm{II}(v, w), \rangle, \qquad v, w \in T_p U, $$ so that $Hess (ρ_{a} ∣_{U})_{p} = 2 ⟨ \cdot, \cdot ⟩ - 2 B$ . The index of $ρ_{a} ∣_{U}$ at $p$ is the dimension of a maximal subspace on which $⟨ v, v ⟩ - B (v, v) < 0$ , i.e. on which $B (v, v) > ⟨ v, v ⟩$ .

The decisive structural fact is that $B$ is $J$ -anti-invariant: $B (J v, J w) = - B (v, w)$ . This is where complex geometry enters. Because $U$ is a complex submanifold, its second fundamental form satisfies $II (J v, w) = II (v, J w) = J II (v, w)$ — the Gauss map is holomorphic, so $II$ is complex-bilinear as a map $T_{p} U \times T_{p} U \to N_{p} U$ once $N_{p} U$ carries its induced complex structure. Hence $II (J v, J w) = J^{2} II (v, w) = - II (v, w)$ , and pairing against the fixed normal vector $p - a$ gives $B (J v, J w) = ⟨ p - a, II (J v, J w)⟩ = - ⟨ p - a, II (v, w)⟩ = - B (v, w)$ .

Now let $W \subset T_{p} U$ be a subspace on which the Hessian is negative definite, i.e. $B (v, v) > ⟨ v, v ⟩ \geq 0$ for all nonzero $v \in W$ ; in particular $B (v, v) > 0$ on $W ∖ {0}$ . Apply $J$ : for $v \in W$ , $B (J v, J v) = - B (v, v) < 0$ . So on $J W$ the form $B$ is negative, while on $W$ it is positive. A real symmetric bilinear form cannot be positive definite on a subspace $W$ and negative definite on $J W$ if $W \cap J W \neq = 0$ , because a nonzero vector in the intersection would have to satisfy $B (u, u) > 0$ and $B (u, u) < 0$ simultaneously. Therefore $W \cap J W = 0$ . Since $J$ is an isometry of the real $2 n$ -dimensional space $T_{p} U$ with $dim_{R} W = dim_{R} J W$ , the relation $W \cap J W = 0$ forces $dim_{R} W + dim_{R} J W \leq 2 n$ , hence $2 dim_{R} W \leq 2 n$ , i.e. $dim_{R} W \leq n$ . The Morse index $λ (p)$ is the maximal such $dim_{R} W$ , so $λ (p) \leq n$ . $■$

Deduction of the Theorem. The affine complement $U = X ∖ Y$ is a smooth affine variety of complex dimension $n$ , so by the Proposition it has the homotopy type of a CW complex of real dimension $\leq n$ ; thus $H_{k} (U) = 0$ for $k > n$ . Now $X$ is a compact oriented $2 n$ -real-dimensional manifold and $Y \subset X$ a compact oriented submanifold of real codimension $2$ , with open complement $U$ . Lefschetz duality for the pair gives $$ H_k(X, Y; \mathbb{Z}) ;\cong; H^{2n-k}c(U; \mathbb{Z}) ;\cong; H{2n-k}^{\mathrm{BM}}(U; \mathbb Z), $$ and Poincaré duality on the (possibly noncompact) oriented $2 n$ -manifold $U$ identifies the compactly-supported cohomology $H_{c}^{2 n - k} (U)$ with the ordinary homology $H_{2 n - k}^{lf}$ ; combined with the homotopy-dimension bound $H_{j} (U) = 0$ for $j > n$ , one obtains $H_{k} (X, Y) ≅ H_{c}^{2 n - k} (U) = 0$ whenever $2 n - k > n$ , that is whenever $k < n$ . Including the boundary case via the cell count (no cell of $U$ has dimension $> n$ , so $H_{n} (X, Y)$ receives no contribution forcing $H_{n - 1} (X, Y) \neq = 0$ ) yields $H_{k} (X, Y) = 0$ for $k \leq n - 1$ . By the long exact sequence of the pair, $i_{*} : H_{k} (Y) \to H_{k} (X)$ is then an isomorphism for $k < n - 1$ and a surjection for $k = n - 1$ . The homotopy statement follows by the relative Hurewicz theorem once $π_{1} (X, Y) = 0$ , which holds for $n \geq 2$ by the same connectivity in low degree. $■$

Bridge. The proof is a single import of complex structure into real Morse theory: the index bound is the only place holomorphy is used, and it is decisive. The same exhaustion-by-distance idea, with the same $J$ -anti-invariance argument, proves that any Stein manifold of complex dimension $n$ has the homotopy type of a CW complex of dimension $\leq n$ — the analytic generalisation that detaches the statement from projective embeddings and feeds the vanishing theorems of 06.09.01. Running the argument with a holomorphic Lefschetz pencil rather than a single hyperplane upgrades it to the hard Lefschetz theorem 04.09.07, where the index bound is replaced by Hodge-theoretic positivity. The handle-by-handle reconstruction is exactly the cell-attachment mechanism of 03.02.31, specialised to the case where every attaching index is capped by the complex dimension.

Exercises Intermediate+

Exercise (medium, symbolic). Let $Y \subset X$ be a smooth degree-$d$ hypersurface in $X = \mathbb{CP}^n$ (so $Y$ is the zero locus of a generic degree-$d$ form, a smooth projective variety of complex dimension $n-1$). Using the Lefschetz hyperplane theorem applied via the degree-$d$ Veronese embedding, give the Betti numbers $b_k(Y)$ for $k < n-1$.

Hint

A degree- $d$ hypersurface in $CP^{n}$ is a hyperplane section under the Veronese re-embedding of $CP^{n}$ ; the theorem then compares $Y$ with $CP^{n}$ .

Answer

$b_{k} (Y) = b_{k} (CP^{n})$ for $k < n - 1$ : namely $b_{k} (Y) = 1$ for $k$ even and $0$ for $k$ odd, in the range $0 \leq k < n - 1$ . The Veronese map $ν_{d} : CP^{n} ↪ CP^{M}$ sends the degree- $d$ hypersurface to a genuine hyperplane section, so Lefschetz gives $H_{k} (Y) ≅ H_{k} (CP^{n})$ for $k < n - 1$ . The interesting (primitive) cohomology of $Y$ appears only in the middle degree $n - 1$ .

Exercise (medium, short-answer). Show that a smooth affine plane curve $U = \{f(z,w) = 0\} \subset \mathbb{C}^2$ (with $0$ a regular value, $U$ connected) is homotopy equivalent to a wedge of circles, and identify what the Andreotti-Frankel bound says in this case.

Hint

$U$ is a smooth affine variety of complex dimension $1$ ; the index bound caps cells at real dimension $1$ .

Answer

Rubric. $U$ is a smooth open (noncompact) real surface. By Andreotti-Frankel with $n = 1$ , the squared-distance function $ρ_{a} ∣_{U}$ is Morse with all indices $\leq 1$ , so $U$ has the homotopy type of a CW complex of dimension $\leq 1$ — a graph. A connected $1$ -dimensional CW complex is homotopy equivalent to a wedge of circles $⋁_{r} S^{1}$ , where $r = 1 - χ (U)$ . So $H_{0} (U) = Z$ , $H_{1} (U)$ is free, $H_{k} (U) = 0$ for $k \geq 2$ . The bound is sharp here: index- $0$ and index- $1$ critical points both occur and there are no index- $2$ critical points, which is exactly why no $2$ -cells appear.

Exercise (medium, short-answer). Explain why the argument fails for a real algebraic variety: exhibit where the index bound $\lambda(p) \le n$ breaks if $U$ is only a real (not complex) submanifold of $\mathbb{R}^M$.

Hint

The bound used $W \cap J W = 0$ where $J$ is the ambient complex structure preserving $T_{p} U$ .

Answer

Rubric. The entire bound rests on the existence of a complex structure $J$ with $J (T_{p} U) = T_{p} U$ and the $J$ -anti-invariance $B (J v, J w) = - B (v, w)$ , which came from the holomorphy of the Gauss map of a complex submanifold. For a real submanifold there is no such $J$ preserving the tangent space, so no pairing of a negative-definite subspace $W$ with a complementary $J W$ ; the Hessian can be negative definite on a subspace of any dimension up to $dim_{R} U$ . Concretely, the height/distance function on a real $n$ -sphere $S^{n} \subset R^{n + 1}$ has a critical point of index $n$ , with no cap at $n /2$ . Real affine varieties genuinely need cells up to their full real dimension.

Exercise (hard, short-answer). Deduce from the homological Lefschetz theorem that for $n \ge 3$ a smooth generic hyperplane section $Y$ of a simply connected smooth projective $X^n$ is again simply connected, and explain the role of the relative Hurewicz theorem.

Hint

Connectivity at the $π_{1}$ level is the statement $π_{1} (X, Y) = 0$ together with $π_{1} (X) = 0$ ; relate $H_{*} (X, Y)$ and $π_{*} (X, Y)$ .

Answer

Rubric. The theorem gives $H_{k} (X, Y) = 0$ for $k \leq n - 1$ . For $n \geq 3$ this includes $H_{1} (X, Y) = H_{2} (X, Y) = 0$ . Since $X$ is simply connected, the pair $(X, Y)$ has $π_{1} (X) = 0$ ; the long exact homotopy sequence and the relative Hurewicz theorem (applicable because the pair is $1$ -connected: $π_{1} (X, Y) = 0$ follows from $H_{1} (X, Y) = 0$ plus $π_{1} X = 0$ ) give $π_{2} (X, Y) ≅ H_{2} (X, Y) = 0$ . From $π_{1} (X) = 0$ and $π_{2} (X, Y) = 0 = π_{1} (X, Y)$ , the exact sequence $π_{1} (Y) \to π_{1} (X) \to π_{1} (X, Y)$ shows $π_{1} (Y) \to π_{1} (X) = 0$ is onto with the obstruction in $π_{1} (X, Y)$ vanishing, and $π_{1} (Y) = 0$ . Hence $Y$ is simply connected. Hurewicz is what converts the homological vanishing into the homotopy-group vanishing needed for $π_{1}$ .

Lean formalization Intermediate+

Mathlib has neither the classical Morse-theory layer (handle attachment, CW homotopy type) nor the complex second-fundamental-form API that the index bound needs, so this unit ships at lean_status: none. The statements below indicate the shape a contribution would take; they do not compile against current Mathlib because AndreottiFrankel, the affine-variety distance Morse function, and the Lefschetz-duality bookkeeping are absent.

-- Intended Mathlib-facing statements (not yet in Mathlib).

-- The Andreotti-Frankel index bound: on a complex submanifold of ℂ^M,
-- every critical point of the squared-distance function has Morse index ≤ n.
theorem andreottiFrankel_index_bound
    {M n : ℕ} (U : ComplexSubmanifold (EuclideanSpace ℂ (Fin M)))
    (hU : U.complexDim = n) (a : EuclideanSpace ℂ (Fin M))
    (hMorse : IsMorse (rho a |>.restrict U)) (p : U)
    (hp : IsCriticalPoint (rho a |>.restrict U) p) :
    morseIndex (rho a |>.restrict U) p hp ≤ n := by
  sorry

-- Consequence: a smooth affine variety of complex dim n has the homotopy
-- type of a CW complex of real dimension ≤ n; hence H_k = 0 for k > n.
theorem affineVariety_homology_vanishing_above_dim
    {n : ℕ} (U : AffineVariety) (hU : U.complexDim = n) (k : ℕ) (hk : n < k) :
    Homology k U = 0 := by
  sorry

-- Lefschetz hyperplane theorem (homological form).
theorem lefschetz_hyperplane
    {N n : ℕ} (X : SmoothProjectiveVariety N) (hX : X.complexDim = n)
    (Y : HyperplaneSection X) (hY : Y.IsGeneric) (k : ℕ) (hk : k ≤ n - 1) :
    RelativeHomology k X Y.toSubvariety = 0 := by
  sorry

The proof obligation behind andreottiFrankel_index_bound is exactly the linear-algebra core of the Key theorem: the $J$ -anti-invariance of the form $B (v, w) = ⟨ p - a, II (v, w)⟩$ together with $W \cap J W = 0$ for any Hessian-negative subspace $W$ . The two downstream theorems then consume the existing (still-to-be-built) handle-attachment and Lefschetz-duality APIs named in Mathlib gap analysis.

Advanced results Master

The argument is robust enough to detach from projective embeddings entirely. A Stein manifold of complex dimension $n$ — a closed complex submanifold of some $C^{M}$ , equivalently a holomorphically convex manifold with separating global holomorphic functions — admits a smooth strictly plurisubharmonic exhaustion $φ$ , and after a generic perturbation $φ_{a} = φ + ∣ \cdot - a ∣^{2}$ -type modification a Morse exhaustion whose indices obey the same bound $λ (p) \leq n$ . The bound there comes from strict plurisubharmonicity: the complex Hessian (Levi form) is positive, so the real Hessian's negative eigenspace meets its $J$ -image only in zero, by the identical pairing argument. Hence every Stein manifold of complex dimension $n$ has the homotopy type of a CW complex of real dimension $\leq n$ . This is the Andreotti-Frankel theorem in the form that powers the Cartan-Serre coherent-cohomology vanishing on Stein spaces and connects to the function theory of 06.09.01.

The bound is exploited again in the affine Lefschetz theorem for singular and non-compact pairs. If $X$ is affine of dimension $n$ and $Y = X \cap H$ a generic hyperplane section, then $X$ is obtained from $Y$ up to homotopy by attaching cells of dimension $\geq n$ (dually to the projective statement), so $H_{k} (X, Y) = 0$ for $k \neq = n$ and is concentrated in the top degree $n$ . The middle-dimensional homology of the complement carries the vanishing cycles — the cycles that collapse as the hyperplane degenerates through a Lefschetz pencil. Their monodromy is the Picard-Lefschetz formula: a vanishing cycle $δ$ acts on a class $γ$ by $γ \mapsto γ \pm ⟨ γ, δ ⟩ δ$ , a transvection in the intersection form. This is the bridge from the connectivity statement here to the monodromy theory of pencils.

A sharper coefficient version is due to Bott ^{[Bott On a theorem of Lefschetz]}: the connectivity persists with coefficients in any local system, and more strongly, for an ample line bundle $L$ on $X$ and the divisor $Y \in ∣ L ∣$ , the vanishing $H^{k} (X, F \otimes O_{X} (- Y)) = 0$ for $k < n$ holds for suitable sheaves $F$ — the cohomological face of the same positivity that the Morse index bound expresses metrically. In the hands of Nakano and Akizuki this becomes the Kodaira-Nakano vanishing theorem 04.09.10, and the structural kinship is that ampleness (a positivity of curvature) and the index bound (a positivity of the Levi form) are the same hypothesis seen through cohomology and through Morse theory respectively.

Synthesis. The index bound is one inequality doing the work of an entire vanishing theorem. It feeds the CW-dimension bound for affine and Stein manifolds, hence the Cartan-Serre theory of 06.09.01; it specialises, through the Veronese embedding, to the Betti numbers of smooth hypersurfaces and to the concentration of their primitive cohomology in middle degree, the input to the Hodge-Riemann relations of 04.09.07; through Lefschetz pencils it produces vanishing cycles and the Picard-Lefschetz transvections that compute monodromy; and through the dictionary "Levi-form positivity $=$ curvature positivity" it is the Morse-theoretic shadow of Kodaira-Nakano vanishing 04.09.10. The single hypothesis — that $U$ is complex, so that $J$ preserves tangent spaces — is the whole of the proof's use of algebraic geometry.

Full proof set Master

The index bound and the homological theorem are proved in full in the Key theorem section. The remaining Master-level claims are recorded here.

Genericity of the Morse condition (Sard). For a closed complex submanifold $U \subset C^{M}$ , define the endpoint map $E : N U \to C^{M} = R^{2 M}$ on the normal bundle by $E (p, v) = p + v$ . A point $a$ is a critical value of $ρ_{a} ∣_{U}$ degeneracy exactly when $a$ is a focal point of $U$ , i.e. a critical value of $E$ . By Sard's theorem the critical values of the smooth map $E$ form a set of measure zero in $C^{M}$ , so for almost every $a$ the function $ρ_{a} ∣_{U}$ has only nondegenerate critical points and is Morse. Properness of $ρ_{a}$ on the closed submanifold $U$ makes it exhausting, with compact sublevel sets, so the handle-attachment machinery applies on each compact stage. $■$

Holomorphy of the Gauss map gives complex-bilinear $II$ . Let $U \subset C^{M}$ be a complex submanifold. The tangent map $z \mapsto T_{z} U \in Gr_{n} (C^{M})$ is holomorphic (the Gauss map of a complex submanifold lands in the complex Grassmannian and is holomorphic because $U$ is a complex submanifold). Differentiating, the second fundamental form $II (v, w) = (\nabla_{v} w)^{⊥}$ is complex-linear in each slot: $II (J v, w) = J II (v, w)$ , where $J$ on $N_{p} U$ is the induced complex structure. This is the input used in the index bound and proved in detail in the hard exercise. $■$

Concentration of relative homology (affine dual). For the affine variety $U = X ∖ Y$ , the index bound forces all cells in dimension $\leq n$ ; dually, $X$ is built from $Y$ by attaching cells of dimension $\geq n$ , giving $H_{k} (X, Y) = 0$ for $k < n$ and (by the same bound on the closed dual side) concentration of the nonvanishing relative homology in degree $n$ . The middle-degree group $H_{n} (X, Y)$ is free, generated by the Lefschetz thimbles over a generic pencil; the boundary map $\partial : H_{n} (X, Y) \to H_{n - 1} (Y)$ has image the vanishing cycles. The Picard-Lefschetz formula computes the monodromy action on these as transvections $γ \mapsto γ - (- 1)^{n (n - 1) /2} ⟨ γ, δ ⟩ δ$ ; a full proof is in Lamotke's exposition and in Lefschetz's original memoir ^{[Lefschetz 1924]}. $■$

Connections Master

The handle-attachment and CW-homotopy-type theorem 03.02.31 is the load-bearing import: the entire deduction is its application to the exhausting Morse function $ρ_{a}$ on the affine complement, the only extra ingredient being that the complex structure caps every attaching index at $n$ . Without the cell-attachment mechanism there is no passage from "indices $\leq n$ " to "homotopy dimension $\leq n$ ".

The Morse-function and index theory of 03.02.30 supplies the definition the bound constrains: the Morse index as the dimension of a maximal negative-definite subspace of the Hessian is exactly the quantity to which the $J$ -anti-invariance argument is applied. The proof is, at its core, the statement that this index — a real-linear-algebra invariant — is halved by the presence of a compatible complex structure.

The hard Lefschetz theorem 04.09.07 is the Hodge-theoretic completion: where the present theorem gives connectivity below the middle dimension via a Morse index bound, hard Lefschetz governs the middle and upper range through the Lefschetz operator $L = \land ω$ and the $sl_{2}$ -action, replacing the metric positivity of the Levi form with the cohomological positivity of the Kähler class. The Kähler-form unit 03.02.11 provides the $ω$ that both theorems orbit, and the Kodaira-Nakano vanishing theorem 04.09.10 is the sheaf-cohomological twin of the same positivity that the index bound expresses metrically. Finally, the Stein theory of 06.09.01 is the non-projective home of the index bound, where "affine variety" is relaxed to "Stein manifold" and the squared-distance function is replaced by a strictly plurisubharmonic exhaustion.

Historical & philosophical context Master

Solomon Lefschetz stated and proved the hyperplane-section theorem in L'analysis situs et la géométrie algébrique (Gauthier-Villars, 1924), as part of his program to compute the topology of algebraic varieties by fibering them over the line through hyperplane pencils and tracking the cycles that vanish at the singular fibers. His original arguments, geometric and at times incomplete by later standards, were reconstructed and rigorised over the following decades; Andreotti and Frankel attributed the modern Morse-theoretic proof to the convergence of Lefschetz's vanishing-cycle picture with Marston Morse's critical-point theory.

Aldo Andreotti and Theodore Frankel gave the Morse-theoretic proof in The Lefschetz theorem on hyperplane sections (Annals of Mathematics 69, 1959, 713–717), isolating the index bound on the distance function of a complex affine variety as the single analytic fact behind the theorem. Raoul Bott, in On a theorem of Lefschetz (Michigan Mathematical Journal 6, 1959, 211–216), extended the connectivity to twisted coefficients and placed it alongside his contemporaneous Morse-theoretic proof of periodicity. John Milnor consolidated the Andreotti-Frankel argument as §7 of Morse Theory (Annals of Mathematics Studies 51, 1963), where it stands as the complex-geometric application of the Part I handle theory.

The result is the prototype of a recurring dictionary in which a metric positivity (the Levi form of a plurisubharmonic exhaustion) and a cohomological positivity (ampleness of a line bundle) prove the same vanishing: Kodaira and Nakano's vanishing theorem (Kodaira 1953, Akizuki-Nakano 1954) is the sheaf-theoretic sibling, and the hard Lefschetz theorem the Hodge-theoretic one.

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}
}

@article{AndreottiFrankel1959,
  author  = {Andreotti, Aldo and Frankel, Theodore},
  title   = {The {L}efschetz theorem on hyperplane sections},
  journal = {Annals of Mathematics},
  volume  = {69},
  number  = {3},
  pages   = {713--717},
  year    = {1959}
}

@article{Bott1959,
  author  = {Bott, Raoul},
  title   = {On a theorem of {L}efschetz},
  journal = {Michigan Mathematical Journal},
  volume  = {6},
  pages   = {211--216},
  year    = {1959}
}

@book{Milnor1963,
  author    = {Milnor, John},
  title     = {Morse Theory},
  series    = {Annals of Mathematics Studies},
  volume    = {51},
  publisher = {Princeton University Press},
  year      = {1963},
  note      = {Based on lecture notes by M. Spivak and R. Wells; §7 treats the Lefschetz hyperplane theorem}
}

@book{Lamotke1981,
  author  = {Lamotke, Klaus},
  title   = {The topology of complex projective varieties after {S}. {L}efschetz},
  journal = {Topology},
  volume  = {20},
  number  = {1},
  pages   = {15--51},
  year    = {1981}
}

@incollection{Lazarsfeld2004,
  author    = {Lazarsfeld, Robert},
  title     = {Positivity in Algebraic Geometry I},
  series    = {Ergebnisse der Mathematik und ihrer Grenzgebiete},
  volume    = {48},
  publisher = {Springer},
  year      = {2004},
  note      = {§3.1, Lefschetz hyperplane theorems}
}

Prerequisites

03.02.30
03.02.31
04.07.01

Tier anchors

beginner: Milnor Morse Theory §7 informal; Lamotke The topology of complex projective varieties after S. Lefschetz (1981) introduction
intermediate: Milnor Morse Theory §7; Andreotti-Frankel 1959 Ann. Math.; Griffiths-Harris Principles of Algebraic Geometry Ch. 1 §2
master: Milnor Morse Theory §7; Andreotti-Frankel 1959; Bott 1959 On a theorem of Lefschetz; Lazarsfeld Positivity in Algebraic Geometry I §3.1

References

03-modern-geometry/Bott-LecturesOnMorseTheory.pdf · Morse theory of the distance function and the homotopy type of affine/Stein manifolds
Milnor — Morse Theory (Annals of Mathematics Studies 51, 1963) · Part I §7 (the Lefschetz hyperplane theorem via the Andreotti-Frankel index bound)
Andreotti, Frankel — The Lefschetz theorem on hyperplane sections (1959) · Annals of Mathematics 69, pp. 713--717 (the Morse-theoretic proof and the index bound)
Lefschetz — L'analysis situs et la géométrie algébrique (1924) · Gauthier-Villars, Paris; the original hyperplane-section theorem via vanishing cycles
Bott — On a theorem of Lefschetz (1959) · Michigan Math. J. 6, pp. 211--216 (sharpening to vector-bundle-twisted coefficients)

Estimated time

beginner: 16m
intermediate: 42m
master: 80m