Overview

  • When are only finitely many fibers F−λ reducible?
  • How does a composition F=h(H) affect its fibers?
  • What does Bertini’s theorem say over infinite and finite fields?

Fibers and composition

A nonconstant polynomial F∈C[x1,…,xr]F\in\mathbb C[x_1,\ldots,x_r], with r≥2r\ge2, defines a map to the affine line. The fiber over λ\lambda is the hypersurface F−λ=0F-\lambda=0. Call FF composite if F=h(H)F=h(H) for a polynomial H∈C[x]H\in\mathbb C[\mathbf x] and a univariate polynomial hh of degree at least two.

If F=h(H)F=h(H), then every F−λF-\lambda factors over C\mathbb C: factor h(T)−λh(T)-\lambda into linear factors and substitute HH. Repeated factors count as reducibility. Thus composition explains a whole family of reducible fibers, rather than a single exceptional value.

The converse holds: if FF is not composite, all but finitely many F−λF-\lambda are irreducible. Equivalently, F−tF-t is irreducible over C(t)‾\overline{\mathbb C(t)}.

References 1

The generic fiber and its field

For every nonconstant FF, the polynomial F−tF-t is already irreducible over C(t)\mathbb C(t): the quotient C[x,t]/(F−t)\mathbb C[\mathbf x,t]/(F-t) is C[x]\mathbb C[\mathbf x], and localization in the nonzero polynomials in tt remains a domain. This ordinary irreducibility does not imply that general complex fibers are irreducible.

Example

Take F=(xy)2F=(xy)^2. The polynomial (xy)2−t(xy)^2-t is irreducible over C(t)\mathbb C(t), but over its algebraic closure it factors as (xy−t)(xy+t)(xy-\sqrt t)(xy+\sqrt t). Every complex fiber is reducible, and the fiber over zero is nonreduced. In contrast, xy−λxy-\lambda is irreducible for every λ≠0\lambda\ne0, by Gauss’s lemma applied in yy.

References 1

The spectrum and Stein’s bound

The spectrum of FF is the set of λ\lambda for which F−λF-\lambda is reducible. For a non-composite F∈C[x,y]F\in\mathbb C[x,y] of total degree dd, let rλr_\lambda be the number of distinct irreducible factors of F−λF-\lambda. Stein’s theorem gives

∑λ∈C(rλ−1)≤d−1.\sum_{\lambda\in\mathbb C}(r_\lambda-1)\le d-1.

In particular, there are at most d−1d-1 reducible fibers. For F=xyF=xy, the only such value is zero, where xyxy has two components; equality holds.

Lorenzini extended the theorem to algebraically closed fields of arbitrary characteristic and refined the bound. Bodin–Dèbes–Najib study how the spectrum changes under specialization and reduction. For a rational function P/QP/Q, the corresponding family is P−λQP-\lambda Q, with a separate fiber at infinity; its bounds and hypotheses differ from the polynomial case. Benelmekki–Najib give a later proof of a rational-function bound using Lüroth’s theorem.

References 4

General linear sections

Bertini’s theorem concerns a different variation: intersect a fixed variety with a varying hyperplane. Over an algebraically closed field, a smooth irreducible projective variety of dimension at least two has a smooth irreducible general hyperplane section. Here “general” means that the hyperplanes lie in a nonempty Zariski open subset of the parameter space.

The dimension assumption matters. A general hyperplane meets a projective curve in finitely many points, and this intersection need not be irreducible. The theorem also concerns varying hyperplanes: it does not say that every section in a chosen pencil is irreducible.

There are quantitative versions. Poonen–Slavov bound the dimension of the exceptional hyperplanes for suitable morphisms to projective space. Kmentt–Shute extend the dimension estimate to linear sections of higher codimension.

References 3

Sections over finite fields

Over a finite field, a nonempty open subset of the parameter space need not contain a rational point. The usual existence argument for a general hyperplane therefore does not apply directly.

Charles–Poonen prove the following replacement. Let X⊆PFqNX\subseteq\mathbb P^N_{\mathbb F_q} be geometrically irreducible with dim⁡X≥2\dim X\ge2. Among all degree-dd homogeneous polynomials over Fq\mathbb F_q, the proportion whose hypersurfaces meet XX in a geometrically irreducible scheme tends to one as d→∞d\to\infty. This is a statement about geometric irreducibility; reducedness is a separate property.

The published erratum corrects part of the proof and leaves the main results unchanged.

References 2

Related topics

References

  1. Indecomposable polynomials and their spectrum

    Bodin, Arnaud and Dèbes, Pierre and Najib, Salah. Acta Arithmetica, 139(1), 79–100, 2009.

    Read the source
  2. The total reducibility order of a polynomial in two variables

    Stein, Yosef. Israel Journal of Mathematics, 68, 109–122, 1989.

    Read the source
  3. Reducibility of polynomials in two variables

    Lorenzini, Dino. Journal of Algebra, 156(1), 65–75, 1993.

    Read the source
  4. A note on the spectrum of a rational function

    Benelmekki, Mohamed and Najib, Salah. Annales mathématiques Blaise Pascal, 30(2), 107–114, 2023.

    Read the source
  5. Bertini theorems

    The Stacks Project Authors. The Stacks Project, 2026.

    Read the source
  6. The Exceptional Locus in the Bertini Irreducibility Theorem for a Morphism

    Poonen, Bjorn and Slavov, Kaloyan. International Mathematics Research Notices, 2022(6), 4503–4513, 2022.

    Read the source
  7. The Bertini irreducibility theorem for higher codimensional slices

    Kmentt, Philip and Shute, Alec. Finite Fields and Their Applications, 83, 102085, 2022.

    Read the source
  8. Bertini irreducibility theorems over finite fields

    Charles, François and Poonen, Bjorn. Journal of the American Mathematical Society, 29(1), 81–94, 2016.

    Read the source
  9. Erratum to “Bertini irreducibility theorems over finite fields”

    Charles, François and Poonen, Bjorn. Journal of the American Mathematical Society, 32(2), 605–607, 2019.

    Read the source