TheoremComplete

Bertini's Theorem

Bertini's theorem says that "a general hyperplane section of a smooth variety is smooth." It is the fundamental tool for inductive arguments in algebraic geometry β€” reducing the study of varieties to lower-dimensional ones by cutting with hyperplanes.


Statement

Theorem1.3Bertini's Theorem

Let XβŠ†PnX \subseteq \mathbb{P}^n be a nonsingular projective variety over an algebraically closed field kk of characteristic 00. Then for a general hyperplane H∈(Pn)βˆ—H \in (\mathbb{P}^n)^* (meaning for all HH in a dense open subset of the dual projective space), the intersection X∩HX \cap H is:

  1. Either empty, or
  2. A nonsingular variety of dimension dim⁑Xβˆ’1\dim X - 1.
RemarkGeneral = generic

"General" means "for all hyperplanes in a dense open subset of the parameter space (Pn)βˆ—(\mathbb{P}^n)^*." There may be finitely many (or a lower-dimensional family of) bad hyperplanes, but "most" hyperplanes give smooth sections. The precise statement is that the set of bad hyperplanes is contained in a proper closed subset of (Pn)βˆ—(\mathbb{P}^n)^*.

RemarkCharacteristic p caveat

Bertini's theorem can fail in characteristic p>0p > 0. For example, consider the Frobenius morphism and inseparable maps. The correct statement in positive characteristic involves separability conditions. The version above (char 00) is the one in Hartshorne II.8.18.


Examples

ExampleHyperplane section of a smooth surface

Let X=V(x02+x12+x22+x32)βŠ†P3X = V(x_0^2 + x_1^2 + x_2^2 + x_3^2) \subseteq \mathbb{P}^3 (a smooth quadric surface, β‰…P1Γ—P1\cong \mathbb{P}^1 \times \mathbb{P}^1). A general hyperplane H=V(a0x0+β‹―+a3x3)H = V(a_0 x_0 + \cdots + a_3 x_3) meets XX in a smooth curve of degree 2 in Hβ‰…P2H \cong \mathbb{P}^2, i.e., a smooth conic β‰…P1\cong \mathbb{P}^1.

A special hyperplane like H=V(x0)H = V(x_0) gives V(x12+x22+x32,x0)V(x_1^2 + x_2^2 + x_3^2, x_0), which is still a smooth conic. In fact, for this particular quadric, every hyperplane section is smooth (or empty) because XX itself is smooth and the dual variety has codimension >0> 0.

But if XX were a quadric cone V(x02+x12βˆ’x22)V(x_0^2 + x_1^2 - x_2^2) (singular at [0:0:0:1][0:0:0:1]), then H=V(x3)H = V(x_3) gives a smooth section, while H=V(x2βˆ’x0)H = V(x_2 - x_0) passes through the vertex and gives a singular section.

ExampleHyperplane section of a curve = points

A smooth curve CβŠ†PnC \subseteq \mathbb{P}^n of degree dd meets a general hyperplane in dd distinct points (smooth = multiplicity 1 at each). Bertini says these dd points are all distinct (no tangencies with a general hyperplane).

For the twisted cubic CβŠ†P3C \subseteq \mathbb{P}^3 (deg⁑C=3\deg C = 3): a general plane meets CC in 3 distinct points.

ExampleBertini for linear systems

The theorem generalizes: let ∣D∣|D| be a linear system (family of divisors) on a smooth variety XX. If ∣D∣|D| is base-point free (no point lies in every member), then a general member of ∣D∣|D| is smooth.

For X=P2X = \mathbb{P}^2 and ∣D∣=∣O(d)∣|D| = |\mathcal{O}(d)| (degree-dd curves), a general degree-dd curve in P2\mathbb{P}^2 is smooth. This is "obvious" for d=1d = 1 (all lines are smooth) and non-trivial for higher dd.

ExampleA pencil of cubics in PΒ²

Consider the pencil (1-parameter family) of cubics {sF+tG=0}[s:t]∈P1\{sF + tG = 0\}_{[s:t] \in \mathbb{P}^1} where FF and GG are two smooth cubics meeting in 9 points (by BΓ©zout). By Bertini, a general member of the pencil is smooth. But finitely many members are singular β€” these correspond to special values of [s:t][s:t].

For the pencil {s(y2zβˆ’x3+xz2)+t(y2zβˆ’x3βˆ’xz2)=0}\{s(y^2z - x^3 + xz^2) + t(y^2z - x^3 - xz^2) = 0\}: a general member is a smooth elliptic curve, but special members degenerate (e.g., a nodal cubic, or a cuspidal cubic, or a reducible cubic).

ExampleConnection to Lefschetz hyperplane theorem

Bertini is the algebraic side of the Lefschetz hyperplane theorem: for a smooth projective variety XβŠ†PnX \subseteq \mathbb{P}^n of dimension β‰₯2\geq 2 and a general hyperplane section Y=X∩HY = X \cap H:

Ο€i(Y)β†’βˆΌΟ€i(X)forΒ i<dim⁑Xβˆ’1.\pi_i(Y) \xrightarrow{\sim} \pi_i(X) \quad \text{for } i < \dim X - 1.

In particular, a smooth surface section of a smooth 3-fold has the same fundamental group. This is a powerful topological tool, and Bertini's smoothness is the algebraic prerequisite.

ExampleInduction on dimension via Bertini

Many theorems in algebraic geometry are proved by induction on dimension, with Bertini providing the inductive step. For example:

Claim: A smooth hypersurface X=V(F)βŠ†PnX = V(F) \subseteq \mathbb{P}^n of degree dβ‰₯2d \geq 2 is connected (for nβ‰₯2n \geq 2).

Proof sketch: By induction on nn. A general hyperplane section X∩HX \cap H is smooth of dimension dim⁑Xβˆ’1β‰₯1\dim X - 1 \geq 1 (by Bertini). If n=2n = 2: XX is a smooth curve in P2\mathbb{P}^2, which is connected. For nβ‰₯3n \geq 3: X∩HX \cap H is connected by induction, and Lefschetz says Ο€0(X∩H)β†’Ο€0(X)\pi_0(X \cap H) \to \pi_0(X) is surjective, so XX is connected.


Variants and generalizations

RemarkOther Bertini-type results
  • Bertini for singularities: If XX is normal with isolated singularities, a general hyperplane section is also normal with isolated singularities.
  • Generic smoothness (char 0): For a dominant morphism f:Xβ†’Yf: X \to Y of varieties in characteristic 0, the fiber fβˆ’1(y)f^{-1}(y) is smooth for yy in a dense open subset of YY.
  • Bertini with multiplicities: If XX has singularities of multiplicity ≀m\leq m, a general hyperplane section also has singularities of multiplicity ≀m\leq m.

Bertini, together with the Lefschetz hyperplane theorem, is one of the main reasons algebraic geometers can reduce problems to the study of curves and surfaces.