Overview

  • When do critical points determine the Galois group of f(x)βˆ’T?
  • Why do simple critical values give transpositions?
  • Why is the generic group symmetric for xβΏβˆ’xβˆ’T?
  • How can the group change after specialization?

Morse polynomials

Let KK have characteristic zero and let f∈K[x]f\in K[x] have degree nβ‰₯2n\ge2. A critical point is a root cc of fβ€²f'; its critical value is f(c)f(c). The polynomial is called Morse if it has nβˆ’1n-1 distinct critical points and their critical values are pairwise distinct.

Both conditions matter. For f=x4βˆ’2x2f=x^4-2x^2, the critical points 0,1,βˆ’10,1,-1 are simple, but f(1)=f(βˆ’1)=βˆ’1f(1)=f(-1)=-1. Moreover, f=(x2)2βˆ’2x2f=(x^2)^2-2x^2 is a composition of two quadratic polynomials. This decomposition gives a block system on the four roots of f(x)βˆ’Tf(x)-T.

The geometric Galois group is taken over Kβ€Ύ(T)\overline K(T); the arithmetic Galois group is taken over K(T)K(T). The theorem below determines both.

References 1
  • Serre2007Topics Β· Β§4.4, Theorems 4.4.1 and 4.4.5, pp. 39–41Morse hypotheses and the geometric construction; the quartic boundary example is computed here.Source

The Morse criterion

Theorem

If f∈K[x]f\in K[x] is Morse and char⁑K=0\operatorname{char}K=0, then the geometric and arithmetic Galois groups of f(x)βˆ’Tf(x)-T are both SnS_n, where n=deg⁑fn=\deg f. Its splitting field over K(T)K(T) is regular over KK.

Once the geometric group is SnS_n, the arithmetic assertion follows: it contains the geometric group and is itself a subgroup of SnS_n. Equality of the groups means that the splitting field introduces no new algebraic constants.

References 1
  • Serre2007Topics Β· Theorems 4.4.1 and 4.4.5The general theorem is applied after base change to the algebraic closure; arithmetic equality and regularity follow.Source

Proof

Work over Kβ€Ύ\overline K. By Gauss’s lemma, a factorization of f(x)βˆ’Tf(x)-T over Kβ€Ύ(T)\overline K(T) would give one in Kβ€Ύ[x,T]\overline K[x,T]. Its degree in TT is one, so one factor is independent of TT. That factor divides the coefficient βˆ’1-1 of TT, and is therefore a unit. Thus f(x)βˆ’Tf(x)-T is irreducible over Kβ€Ύ(T)\overline K(T), and its group GG is transitive.

At a simple critical point cc, the expansion has the form f(c)+a(xβˆ’c)2+β‹―f(c)+a(x-c)^2+\cdots, with aβ‰ 0a\ne0. The ramification index at cc is two. Since no other critical point has value f(c)f(c), all other points of this fiber are unramified. An inertia generator therefore exchanges exactly two roots and fixes the rest: it is a transposition.

Let HH be the normal subgroup generated by inertia at the finite branch values. The quotient of the Galois closure by HH is a connected degree-dd cover of P1\mathbb P^1, unramified except possibly at infinity. If rr points lie above infinity, Riemann–Hurwitz gives

2gβˆ’2=βˆ’2d+(dβˆ’r)=βˆ’dβˆ’r.2g-2=-2d+(d-r)=-d-r.

Since gβ‰₯0g\ge0 and d,rβ‰₯1d,r\ge1, we get d=r=1d=r=1. Thus H=GH=G: the transpositions generate the whole group.

Make a graph whose vertices are the roots and whose edges are the generating transpositions. Transitivity says that this graph is connected. Edge transpositions of a connected graph generate SnS_n, so G=SnG=S_n.

References 1
  • Serre2007Topics Β· Theorem 4.4.5 and Proposition 4.4.6, p. 41Inertia generation via Riemann–Hurwitz. The graph argument expands the transposition step.Source

Connections and evidence

Classical

Morse geometry and Hilbert specialization require two different inputs

Simple critical points and distinct critical values yield regular SnS_n-monodromy; the Hilbert property then supplies infinitely many specializations with the same group.

Scope. Let KK be a number field, f∈K[X]f\in K[X] monic of degree nβ‰₯2n\ge2, and tt transcendental. Critical points are taken over Kβ€Ύ\overline K.

Morse geometry and Hilbert specialization require two different inputsSimple critical points and distinct critical values yield regular \(S_n\)-monodromy; the Hilbert property then supplies infinitely many specializations with the same group. Connection 1: All critical points simple is an input to AND. The derivative has exactly \(n-1\) distinct roots over \(\overline K\). Connection 2: All critical values distinct is an input to AND. The images of those critical points are pairwise distinct. Connection 3: AND leads to Transposition inertia generates. Each finite branch inertia group is generated by a transposition, and their conjugates generate geometric monodromy. Connection 4: Transposition inertia generates leads to Regular generic group Sβ‚™. The polynomial cover is transitive, so generation by transpositions yields geometric and arithmetic group \(S_n\), with constant field \(K\). Connection 5: Regular generic group Sβ‚™ is an input to AND. The splitting field is regular over the coefficient field and has full symmetric group. Connection 6: Number field is Hilbertian is an input to AND. The affine line over a number field has the Hilbert property. Connection 7: AND leads to Sβ‚™ outside a thin set. Outside a thin subset of \(K\), the polynomial \(f(X)-a\) is irreducible with Galois group \(S_n\). Connection 8: AND leads to Infinitely many disjoint Sβ‚™ fields. Regularity and the Hilbert property jointly allow specializations linearly disjoint from any prescribed finite extension, and hence infinitely many linearly disjoint splitting fields.All critical points simple β€” related articleAll critical points simpleAll critical values distinct β€” related articleAll critical values distinctANDANDTransposition inertia generates β€” related articleTransposition inertiageneratesRegular generic group Sβ‚™ β€” related articleRegular generic group Sβ‚™Number field is Hilbertian β€” related articleNumber field is HilbertianANDANDSβ‚™ outside a thin set β€” related articleSβ‚™ outside a thin setInfinitely many disjoint Sβ‚™ fields β€” related articleInfinitely many disjointSβ‚™ fieldsConnection 1: The derivative has exactly \(n-1\) distinct roots over \(\overline K\).First Morse condition[1]Connection 2: The images of those critical points are pairwise distinct.Second Morse condition[2]Connection 3: Each finite branch inertia group is generated by a transposition, and their conjugates generate geometric monodromy.Simple branching [3]Connection 4: The polynomial cover is transitive, so generation by transpositions yields geometric and arithmetic group \(S_n\), with constant field \(K\).Connected transpositionaction [4]Connection 5: The splitting field is regular over the coefficient field and has full symmetric group.Generic Galois input [5]Connection 6: The affine line over a number field has the Hilbert property.Arithmetic input [6]Connection 7: Outside a thin subset of \(K\), the polynomial \(f(X)-a\) is irreducible with Galois group \(S_n\).Hilbert irreducibility [7]Connection 8: Regularity and the Hilbert property jointly allow specializations linearly disjoint from any prescribed finite extension, and hence infinitely many linearly disjoint splitting fields.Regular Hilbertspecialization [8]
Linked nodes open articles; numbered connections open the evidence below.Open full-size diagram

deduction or information flowAND joint hypotheses

A polynomial is Morse when the nβˆ’1n-1 roots of fβ€²f' are simple and have pairwise distinct images. Under the displayed assumptions, the splitting field LL of f(X)βˆ’tf(X)-t satisfies

Gal⁑(L/K(t))=Sn,L∩Kβ€Ύ=K.\operatorname{Gal}(L/K(t))=S_n,\qquad L\cap\overline K=K.

The proof first works geometrically. Each finite branch point contributes a transposition, and these inertia groups and their conjugates generate the transitive monodromy group. Hence that group is SnS_n. Arithmetic monodromy lies between geometric monodromy and SnS_n, so they coincide and regularity follows.

Hilbert irreducibility adds a different input: because KK is a number field, there is a thin set EβŠ‚KE\subset K such that every a∈Kβˆ–Ea\in K\setminus E gives an irreducible f(X)βˆ’af(X)-a with group SnS_n. Regularity further permits infinitely many splitting fields that are linearly disjoint. These conclusions concern suitable specializations; the exceptional set need not be finite.

For example, f=X3βˆ’Xf=X^3-X is Morse. At a=0a=0, its fiber splits completely; at a=1a=1, X3βˆ’Xβˆ’1X^3-X-1 is irreducible by the rational-root test and has nonsquare discriminant βˆ’23-23, so its group is S3S_3. The two fibers illustrate why a generic monodromy theorem alone does not certify a chosen arithmetic fiber.

Inspect all 8 connections and exact references
Evidence for each connection
ConnectionStatement and source
1All critical points simple β†’ ANDConstruction / input

The derivative has exactly nβˆ’1n-1 distinct roots over Kβ€Ύ\overline K.

2All critical values distinct β†’ ANDConstruction / input

The images of those critical points are pairwise distinct.

3AND β†’ Transposition inertia generatesDeduction

Each finite branch inertia group is generated by a transposition, and their conjugates generate geometric monodromy.

4Transposition inertia generates β†’ Regular generic group Sβ‚™Deduction

The polynomial cover is transitive, so generation by transpositions yields geometric and arithmetic group SnS_n, with constant field KK.

5Regular generic group Sβ‚™ β†’ ANDConstruction / input

The splitting field is regular over the coefficient field and has full symmetric group.

6Number field is Hilbertian β†’ ANDConstruction / input

The affine line over a number field has the Hilbert property.

7AND β†’ Sβ‚™ outside a thin setDeduction

Outside a thin subset of KK, the polynomial f(X)βˆ’af(X)-a is irreducible with Galois group SnS_n.

8AND β†’ Infinitely many disjoint Sβ‚™ fieldsDeduction

Regularity and the Hilbert property jointly allow specializations linearly disjoint from any prescribed finite extension, and hence infinitely many linearly disjoint splitting fields.

Sources and scope checked .

Example

Take f(x)=xnβˆ’xf(x)=x^n-x, with nβ‰₯2n\ge2. Its critical points satisfy ncnβˆ’1=1nc^{n-1}=1, so they are nonzero and distinct. At such a point,

f(c)=βˆ’nβˆ’1nc.f(c)=-\frac{n-1}{n}c.

The critical values are therefore distinct. The Morse criterion gives group SnS_n for xnβˆ’xβˆ’Tx^n-x-T in characteristic zero. Over Q\mathbb Q, Hilbert irreducibility then gives infinitely many rational tt for which xnβˆ’xβˆ’tx^n-x-t also has group SnS_n.

References 1
  • Serre2007Topics Β· Β§4.4 and Β§3.3The Morse theorem and specialization framework; the derivative calculation is supplied here.Source

Specializations of a cubic

For F=X3βˆ’3XF=X^3-3X, the critical points are Β±1\pm1, with values βˆ“2\mp2. Thus F(X)βˆ’TF(X)-T has generic group S3S_3. Its discriminant is

Ξ”(X3βˆ’3Xβˆ’T)=27(4βˆ’T2).\Delta(X^3-3X-T)=27(4-T^2).

Examples over Q\mathbb Q

  • t=0t=0: X(X2βˆ’3)X(X^2-3) is reducible; its splitting-field group is C2C_2, fixing the rational root.
  • t=1t=1: X3βˆ’3Xβˆ’1X^3-3X-1 has no rational root and discriminant 8181, so its group is A3=C3A_3=C_3.
  • t=2t=2: (Xβˆ’2)(X+1)2(X-2)(X+1)^2 has a repeated root. This is a branch value, outside the separable-fiber setting.
  • t=3t=3: X3βˆ’3Xβˆ’3X^3-3X-3 is Eisenstein at 33, and its discriminant is βˆ’135-135, a nonsquare. Its group is S3S_3.
References 2

Positive characteristic

In characteristic pp, the polynomial xpβˆ’xx^p-x has derivative βˆ’1-1. The cover xpβˆ’x=Tx^p-x=T is unramified at finite points and has geometric group CpC_p, acting by x↦x+ax\mapsto x+a for a∈Fpa\in\mathbb F_p. It is wildly ramified at infinity, so the tame Riemann–Hurwitz calculation in the proof does not apply. Positive-characteristic versions of the Morse criterion require separate hypotheses.

References 1
  • Serre2007Topics Β· Theorem 4.4.5; remark after Corollary 4.4.7, pp. 41–42Positive-characteristic scope and the Artin–Schreier boundary.Source

Questions

SettledSpecialization and inverse Galois

Do separable specializations retain the generic group?

Scope. The characteristic-zero Morse family XΒ³βˆ’3Xβˆ’T over Q

No. At T=0T=0 the polynomial is reducible, and at T=1T=1 it is irreducible with group C3C_3. Both fibers are separable. Hilbert irreducibility gives many full-group fibers, with possible exceptional specializations.

References 1
  • Serre2007Topics Β· Β§3.3 and Β§4.4Specialization framework; the two counterexamples are calculated in this article.Source
Discuss this question β†’

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. Galois groups of cubics and quartics (not in characteristic 2)

    Conrad, Keith. Author-hosted expository note.

    Read the source