Overview

  • When does specialization preserve irreducibility and the Galois group?
  • How do arithmetic and geometric Galois groups differ?
  • When are the exceptional parameters finite?
  • What changes for powers, units, or algebraic parameters?

From a family to a polynomial

A polynomial F(T,x)F(T,x) describes a family: choosing T=tT=t gives one polynomial F(t,x)F(t,x). Its group over the function field K(T)K(T) is called the generic Galois group. The main question is which values tt preserve this group.

Let KK be a number field and let F∈K(T)[x]F\in K(T)[x] be monic, irreducible, and separable. Write GG for its Galois group on the roots. Away from poles of the coefficients and zeros of the discriminant, the specialized group GtG_t embeds in GG, with the same action on the specialized roots, up to relabelling. A specialization may introduce a factor or reduce the group without introducing repeated roots.

References 1

Arithmetic and geometric groups

Let KK have characteristic zero and let L/K(T)L/K(T) be a finite Galois extension. Its constant field is C=L∩K‾C=L\cap\overline K. The arithmetic group and geometric group fit into

1⟶Gal⁡(LK‾/K‾(T))⟶Gal⁡(L/K(T))⟶Gal⁡(C/K)⟶1.1\longrightarrow \operatorname{Gal}(L\overline K/\overline K(T))\longrightarrow \operatorname{Gal}(L/K(T))\longrightarrow \operatorname{Gal}(C/K)\longrightarrow1.

The first group is naturally isomorphic to Gal⁡(L/C(T))\operatorname{Gal}(L/C(T)). The extension is regular over KK when C=KC=K; in that case the two groups coincide. The exact sequence need not split.

For x3−Tx^3-T over Q(T)\mathbb Q(T), the roots are T3,ζ3T3,ζ32T3\sqrt[3]T,\zeta_3\sqrt[3]T,\zeta_3^2\sqrt[3]T. The arithmetic group is S3S_3, the geometric group is C3C_3, and the constant field is Q(ζ3)\mathbb Q(\zeta_3). For x3−x−Tx^3-x-T, the polynomial is irreducible over Q‾(T)\overline{\mathbb Q}(T), and its discriminant 4−27T24-27T^2 is not a square there. Both groups are S3S_3, so this splitting field is regular.

References 3

Hilbert irreducibility

Theorem

Let KK be a number field and let F∈K(T)[x]F\in K(T)[x] be monic, irreducible, and separable, with Galois group GG. Outside a thin subset of KK, the polynomial F(t,x)F(t,x) is irreducible and has Galois group GG.

On the parameter line, a thin set is contained in the union of a finite set and finitely many images πi(Ci(K))\pi_i(C_i(K)), where CiC_i are geometrically integral curves and πi:Ci→A1\pi_i:C_i\to\mathbb A^1 are dominant maps of degree at least two. For example, rational squares form a thin set through the map u↦u2u\mapsto u^2. Such a set can be infinite.

Hilbert's theorem, proved in 1892, says that KK itself is not thin. Thus infinitely many specializations preserve the generic group. The theorem concerns the arithmetic group over K(T)K(T); regularity is not required.

References 2

Example

Consider F(T,x)=x3−x−TF(T,x)=x^3-x-T, with generic group S3S_3. Three specializations show different possibilities.

  • At t=1t=1, reduction modulo 22 is x3+x+1x^3+x+1, which has no root in F2\mathbb F_2. Thus F(1,x)F(1,x) is irreducible. Its discriminant is −23-23, so its group is S3S_3.
  • If t=u3−ut=u^3-u with u∈Qu\in\mathbb Q, then uu is a root. Conversely, every reducible rational cubic in this family has such a parameter.
  • At t=1/7t=1/7, the primitive integral polynomial 7x3−7x−17x^3-7x-1 again reduces to x3+x+1x^3+x+1 modulo 22. It is irreducible, but the monic cubic has discriminant 169/49=(13/7)2169/49=(13/7)^2, so its group is C3C_3.

Reducibility and a drop in the Galois group are therefore different kinds of exception.

References 2

Finite exceptional sets

The exceptional parameters can sometimes be described by rational points on finitely many curves. If each relevant smooth projective curve has genus at least two, Faltings's theorem makes the exceptional set finite. Curves of genus zero or one require separate analysis.

For generalized Laguerre polynomials, Hajir–Wong prove the following. Fix a number field KK and an integer n≥5n\ge5. For all but finitely many α∈K\alpha\in K, the polynomial Ln(α)L_n^{(\alpha)} is irreducible over KK, and its group contains AnA_n. If n≥10n\ge10, its group equals SnS_n outside a finite set. The degree and the ground field are fixed; the exceptional set need not be effectively known.

The article Exceptional specializations explains the curves, counting bounds, and recent results for algebraic parameters.

References 1

Power parameters

Hilbert irreducibility over all rational parameters does not by itself control a prescribed sequence such as 2n2^n. For X2−tX^2-t, exactly the odd integers nn make X2−2nX^2-2^n irreducible over Q\mathbb Q; their density is 1/21/2.

Theorem (Bary-Soroker–Garzoni–Matei, conditional on GRH)

Let f(t,X)∈Z[t1,…,tr,X]f(\mathbf t,X)\in\mathbb Z[t_1,\ldots,t_r,X], with deg⁡Xf≥1\deg_Xf\ge1, satisfy the pullback condition: for every tuple mi≥1m_i\ge1,

f(t1m1,…,trmr,X) is irreducible in Q‾[t,X].f(t_1^{m_1},\ldots,t_r^{m_r},X)\text{ is irreducible in }\overline{\mathbb Q}[\mathbf t,X].

Fix ai∈Z∖{0,1,−1}a_i\in\mathbb Z\setminus\{0,1,-1\}. Assuming GRH, the proportion of exponent vectors n∈([−N,N]∩Z)r\mathbf n\in([-N,N]\cap\mathbb Z)^r for which f(a1n1,…,arnr,X)f(a_1^{n_1},\ldots,a_r^{n_r},X) is irreducible over Q\mathbb Q tends to one as N→∞N\to\infty.

For X2−tX^2-t, the substitution t=u2t=u^2 produces a factorization, so the pullback condition fails. The theorem concerns irreducibility; preserving the full group also requires control of resolvents.

For one parameter, Dèbes proves a stronger unconditional result: under the same pullback condition, f(an,X)f(a^n,X) is irreducible for all but finitely many integers nn. The proof uses Siegel’s theorem on integral points; it does not determine the exceptional exponents effectively.

References 2
  • BarySorokerGarzoniMatei2025Powers · §1, condition (PB), Theorem 1.1, and the following examplesPublished theorem; GRH and absolute irreducibility of every power pullback are essential.Source
  • Debes1992Powers · Theorem 1.2, p. 144; see also Bary-Soroker–Garzoni–Matei, §1 after Theorem 1.1The original theorem characterizes infinitely many reducible power specializations; geometric (PB) rules out its scaled power-pullback obstruction.Source

Unit parameters

Claimed in a preprint, 3 September 2026. Let π:X→P1\pi:X\to\mathbb P^1 be a nonconstant morphism from a smooth projective irreducible curve over a number field KK, let SS be a finite set of non-archimedean places of KK, and put U=OS×U=\mathcal O_S^\times. Let Red⁡(π)\operatorname{Red}(\pi) be the set of KK-rational parameters with a reducible fiber, together with all KK-rational branch values. Siksek–Stoll state

U∩Red⁡(π)=I∪⋃j=1kγjUrj,∣I∣<∞,γj∈U,rj≥2.U\cap\operatorname{Red}(\pi)=I\cup\bigcup_{j=1}^{k}\gamma_j U^{r_j},\qquad |I|<\infty,\quad \gamma_j\in U,\quad r_j\ge2.

The power cosets are effectively computable; the proof does not make the finite remainder II effective. This describes the exceptional unit parameters more precisely than thinness alone. For x2−Tx^2-T, the exceptional unit parameters are simply the squares U2U^2: if a square in KK is an SS-unit, its square root is also an SS-unit.

References 1
  • SiksekStoll2026 · Theorem 3 and following remarkVersion 1; this is a preprint theorem, not a complete effective algorithm.Source

Finite and infinite families

Hilbert irreducibility also applies to finitely many polynomial families at once: take the union of their thin exceptional sets. For an infinite tower, the statement “each finite level has many good specializations” does not imply that one parameter works at every level. Simultaneous control of all levels requires an additional argument, as in questions about arboreal Galois representations.

References 2

Dynatomic specializations

For fT(x)=x2+Tf_T(x)=x^2+T and n≥1n\ge1, put

Φn(T,x)=∏d∣n(fT∘d(x)−x)μ(n/d).\Phi_n(T,x)=\prod_{d\mid n}(f_T^{\circ d}(x)-x)^{\mu(n/d)}.

Bousch proved that over Q(T)\mathbb Q(T) this polynomial is irreducible and its group is Cn≀SrC_n\wr S_r, where nr=deg⁡xΦnnr=\deg_x\Phi_n. The action must commute with the permutation given by iteration, whose cycles all have length nn. This gives the upper bound Cn≀SrC_n\wr S_r; the generic group attains it. Schleicher gives another treatment through analytic continuation and internal addresses. Let EnE_n be the rational parameters where the specialized group differs from the generic group. Krumm proves EnE_n finite for n=5,6,7,9n=5,6,7,9, whereas it is infinite for n≤4n\le4. These finiteness proofs do not list all exceptional parameters.

References 3

Period four

For every quadratic f∈Q[x]f\in\mathbb Q[x], Krumm determines the Galois group and the degrees of the irreducible factors of its fourth dynatomic polynomial. Thus period four has an explicit classification, while the finiteness question above concerns larger periods. The proof uses subgroup computations and rational points on the corresponding specialization curves.

References 1

Local groups under specialization

Hilbert irreducibility controls the global group. Ramification imposes additional local conditions. Let E/K(T)E/K(T) be a finite Galois extension, regular over a number field KK, and let ti∈Kt_i\in K be a finite branch point. Write Ii◃DiI_i\triangleleft D_i for its inertia and decomposition groups and πi:Di→Di/Ii\pi_i:D_i\to D_i/I_i for the quotient map. The residue extension at tit_i has group Di/IiD_i/I_i.

Let t0∈Kt_0\in K be a nonbranch specialization point. Outside a finite set of primes, suppose t0,tit_0,t_i are integral at p\mathfrak p, a=vp(t0−ti)>0a=v_{\mathfrak p}(t_0-t_i)>0, and gcd⁡(a,∣Ii∣)=1\gcd(a,|I_i|)=1. If Di,pD_{i,\mathfrak p} denotes the decomposition group in the branch-point residue extension, then the decomposition group of the specialization at p\mathfrak p is conjugate to πi−1(Di,p)\pi_i^{-1}(D_{i,\mathfrak p}). This gives a method for prescribing local behavior while seeking specializations with the full global group.

References 1

Computing the two groups

Computations over Q(T)\mathbb Q(T) must identify both the Galois group and its constant field. Fieker–Sutherland give algorithms for these tasks. The constant field determines the quotient of the arithmetic group by the geometric group in the exact sequence above.

References 1

Resolvent curves

Let L/Q(T)L/\mathbb Q(T) be the generic splitting field. For each maximal proper subgroup MM of GG, a defining polynomial for LML^M gives a resolvent curve. Outside finitely many excluded parameters, GtG_t lies in a conjugate of MM exactly when this resolvent has a rational root at tt. This turns group determination into a problem about rational points.

Krumm–Sutherland implement this method and determine all exceptions in several examples. For instance, x6+c6−1x^6+c^6-1, with c∈Qc\in\mathbb Q, is reducible exactly for c∈{0,1,−1}c\in\{0,1,-1\}.

References 1

Questions

Research directionSpecialization and inverse Galois

Can the finite exceptional parameters be determined effectively?

Scope. A specified one-parameter family over a fixed number field

A finiteness theorem need not determine the exceptional parameters effectively. Hajir–Wong discuss this difficulty for their curves; the 2026 unit-specialization theorem also leaves a finite set ineffective. For a given family, the problem reduces to determining rational or integral points on the associated curves.

References 2
Discuss this question →
OpenSpecialization and inverse Galois

Are the exceptional dynatomic specializations finite for every period n>4?

Scope. Φ_n(T,x) for x²+T; specialization T=c∈Q

Krumm proposes finiteness for every n>4n>4 and proves it for 5,6,7,95,6,7,9. Determining each exceptional set explicitly is a further problem.

References 1
Discuss this question →

Related topics

References

  1. Topics in Galois Theory

    Serre, Jean-Pierre. Edition 2. A K Peters/CRC Press, Research Notes in Mathematics 1, 2007.

    Read the source
  2. Generic Polynomials: Constructive Aspects of the Inverse Galois Problem

    Jensen, Christian U. and Ledet, Arne and Yui, Noriko. Mathematical Sciences Research Institute Publications 45, Cambridge University Press, 2002.

    Read the source
  3. Fields and Galois Theory

    Milne, James S. Kea Books, 2022.

    Author’s book page
  4. Galois groups of cubics and quartics (not in characteristic 2)

    Conrad, Keith. Author-hosted expository note.

    Read the source
  5. Ueber die Irreducibilität ganzer rationaler Functionen mit ganzzahligen Coefficienten

    Hilbert, David. Journal für die reine und angewandte Mathematik, 110, 104–129, 1892.

    Read the source
  6. Specializations of one-parameter families of polynomials

    Hajir, Farshid and Wong, Siman. Annales de l’Institut Fourier, 56(4), 1127–1163, 2006.

    Read the source
  7. On the irreducibility of f(2^n,3^m,X) and other such polynomials

    Bary-Soroker, Lior and Garzoni, Daniele and Matei, Vlad. Mathematische Zeitschrift, 310, Article 11, 2025.

    Read the source
  8. On the irreducibility of the polynomials P(t^m, Y)

    Dèbes, Pierre. Journal of Number Theory, 42(2), 141–157, 1992.

    Read the source
  9. Hilbert’s Irreducibility for Gm\mathbb{G}_m

    Siksek, Samir and Stoll, Michael. arXiv preprint, 2026.

    Read the source
  10. Field Arithmetic

    Fried, Michael D. and Jarden, Moshe. Edition 4. Springer, Ergebnisse der Mathematik und ihrer Grenzgebiete, 3. Folge, 11, 2023.

    Read the source
  11. Odoni's conjecture on arboreal Galois representations is false

    Dittmann, Philip and Kadets, Borys. Proceedings of the American Mathematical Society, 150(8), 3335–3343, 2022.

    Read the source
  12. A finiteness theorem for specializations of dynatomic polynomials

    Krumm, David. Algebra & Number Theory, 13(4), 963–993, 2019.

    Read the source
  13. Sur quelques problèmes de dynamique holomorphe

    Bousch, Thierry. Université de Paris-Sud, Orsay, 1992.

    Read the source
  14. Internal Addresses of the Mandelbrot Set and Galois Groups of Polynomials

    Schleicher, Dierk. Arnold Mathematical Journal, 3, 1–35, 2017.

    Read the source
  15. Galois groups in a family of dynatomic polynomials

    Krumm, David. Journal of Number Theory, 187, 469–511, 2018.

    Read the source
  16. On the local behavior of specializations of function field extensions

    König, Joachim and Legrand, François and Neftin, Danny. International Mathematics Research Notices, 2019(9), 2951–2980, 2019.

    Read the source
  17. Computing splitting fields using Galois theory and other Galois constructions

    Fieker, Claus and Sutherland, Nicole. Journal of Symbolic Computation, 116, 243–262, 2023.

    Read the source
  18. Galois groups over rational function fields and explicit Hilbert irreducibility

    Krumm, David and Sutherland, Nicole. Journal of Symbolic Computation, 103, 108–126, 2021.

    Read the source