Overview

  • Which rational factors follow from the generating functions?
  • For which indices does Carlitz’s criterion prove irreducibility?
  • Which Bernoulli and Euler polynomials have repeated roots?
  • How does reflection constrain the Galois group?

Definitions

Define the standard Appell families by

textet−1=∑n≥0Bn(x)tnn!,2extet+1=∑n≥0En(x)tnn!.\frac{te^{xt}}{e^t-1}=\sum_{n\ge0}B_n(x)\frac{t^n}{n!},\qquad \frac{2e^{xt}}{e^t+1}=\sum_{n\ge0}E_n(x)\frac{t^n}{n!}.

The Bernoulli difference identity Bn(x+1)−Bn(x)=nxn−1B_n(x+1)-B_n(x)=nx^{n-1} explains their role in sums of powers. The Euler identity En(x+1)+En(x)=2xnE_n(x+1)+E_n(x)=2x^n is the corresponding alternating relation. Both families satisfy the Appell derivative rule Pn′=nPn−1P_n'=nP_{n-1}. The derivative identity relates multiple roots to common roots of adjacent polynomials.

The notation EnE_n here always denotes the Euler polynomial defined above, not a normalized Eulerian descent polynomial.

References 1

Reflection symmetry

Identity

Bn(1−x)=(−1)nBn(x),En(1−x)=(−1)nEn(x).B_n(1-x)=(-1)^nB_n(x),\qquad E_n(1-x)=(-1)^nE_n(x).

Replacing xx by z+1/2z+1/2 makes each polynomial even or odd according to nn. Hence for odd nn, x−1/2x-1/2 is forced. After removing that factor, the polynomial is a polynomial in (x−1/2)2(x-1/2)^2. Its nonfixed roots come in pairs {α,1−α}\{\alpha,1-\alpha\}. For a separable remaining degree 2m2m, the Galois action embeds in C2≀SmC_2\wr S_m, just as for sign pairs of an even polynomial.

The irreducibility problem and the permutation action depend on which forced and repeated factors have been removed.

References 2

Rational factors

For odd n≥3n\ge3, the Bernoulli values at 00 and 11 vanish, so x(x−1)(x−1/2)x(x-1)(x-1/2) divides Bn(x)B_n(x). For even n≥2n\ge2, the Euler polynomial vanishes at 00 and 11, so x(x−1)x(x-1) divides En(x)E_n(x). These statements follow directly from the generating functions and reflection. Their removal need not leave an irreducible polynomial.

References 3

Carlitz’s irreducibility criterion

Theorem (Carlitz)

Let pp be prime and n=k(p−1)prn=k(p-1)p^r be even, with 1≤k<p1\le k<p and r≥0r\ge0. Then Bn(x)B_n(x) is irreducible over Q\mathbb Q. For p=2p=2, the evenness condition means n=2rn=2^r with r≥1r\ge1.

Proof sketch

Write Bj=Bj(0)B_j=B_j(0). Von Staudt–Clausen says that for even j≥2j\ge2, the denominator of BjB_j contains pp exactly when p−1∣jp-1\mid j, and then only to the first power. The base-pp digits of n=(p−k)pr+(k−1)pr+1n=(p-k)p^r+(k-1)p^{r+1} have sum p−1p-1. If 0<j<n0<j<n and p−1∣jp-1\mid j, Lucas’s congruence gives p∣(nj)p\mid\binom nj: otherwise each digit of jj would be bounded by the corresponding digit of nn, forcing its positive digit sum, a multiple of p−1p-1, to equal p−1p-1, and hence j=nj=n.

Consequently every nonconstant coefficient of pBn(x)pB_n(x) is divisible by pp in Zp\mathbb Z_p, whereas its constant coefficient pBnpB_n is a unit. Its leading coefficient is exactly pp. The exceptional Bernoulli number B1=−1/2B_1=-1/2 causes no problem: for p=2p=2, its coefficient is −n-n, which is even. Thus pxnBn(1/x)px^nB_n(1/x) is Eisenstein over Zp\mathbb Z_p. Taking reciprocals preserves irreducibility.

Example (n=4, p=5)

30B4(x)=30x4−60x3+30x2−1.30B_4(x)=30x^4-60x^3+30x^2-1.

Its reciprocal −x4+30x2−60x+30-x^4+30x^2-60x+30 is Eisenstein at 55. Hence the polynomial is irreducible over Q\mathbb Q.

References 3
  • Carlitz1952BernoulliEulerOriginal irreducibility source. The stated scope is explicitly restated in Kimura’s report; no unchecked original theorem number is claimed.Source
  • Kimura2003Bernoulli · §2, p. 3, theorem attributed to Carlitz (1952)The report uses the reversed coefficient convention for a p-Eisenstein polynomial; the reciprocal formulation above uses the usual convention.Source
  • DLMFBernoulliEuler · §24.10(i), equation (24.10.1)Bernoulli denominators; the displayed proof supplies the digit argument explicitly.Source

Examples

Example

B5(x)=x(x−1)(x−12)(x2−x−13).B_5(x)=x(x-1)(x-\tfrac12)(x^2-x-\tfrac13).

After removing the three rational factors, the quadratic has discriminant 7/37/3, so its splitting field is Q(21)\mathbb Q(\sqrt{21}).

Example

E5(x)=(x−12)(x2−x−1)2.E_5(x)=(x-\tfrac12)(x^2-x-1)^2.

The residual polynomial is a square. Its distinct roots generate only Q(5)\mathbb Q(\sqrt5). Thus removing the rational factor alone need not leave an irreducible or separable polynomial. The splitting field and its Galois action are defined using the distinct roots.

References 2

Multiple roots

Theorem (Brillhart; Dilcher)

Every Bn(x)B_n(x), n≥1n\ge1, is squarefree. Among the Euler polynomials, only E5(x)E_5(x) has multiple roots; its two irrational roots each have multiplicity two.

Brillhart proved the odd Bernoulli case and restricted any possible even-degree multiple root to a root of x2−x−bx^2-x-b, with bb a positive odd integer. Dilcher’s Theorem 2 eliminates these remaining possibilities. The method combines centered expansions, Bernoulli denominator congruences, and the requirement that a multiple root also annihilate the adjacent Appell polynomial. Brillhart’s Euler classification is also recorded in Dilcher’s §4.

References 3

Fixed-point-free automorphisms

Theorem (Patel–Siksek)

For every even n≥2n\ge2, Gal⁡(Bn/Q)\operatorname{Gal}(B_n/\mathbb Q) contains an element fixing none of the roots. Also Bn(x)B_n(x) has no root in Q2\mathbb Q_2.

The proof uses the 22-adic Newton polygon, the pairing α↔1−α\alpha\leftrightarrow1-\alpha, and a decomposition-group argument. Chebotarev then supplies a positive density of primes at which BnB_n has no root. This yields obstructions to equations involving sums of consecutive powers. A permutation without fixed points does not by itself imply that the entire Galois action is transitive.

References 1
  • PatelSiksek2017Powers · Proposition 3.1, Theorem 4, and §§3.2, 5The group and local-root conclusions are unconditional, independent of the conjectured irreducibility of every even Bernoulli polynomial.Source

Normalized polynomials

After the rational factors are removed, centering at 1/21/2 gives an even polynomial. Writing it as Q((x−1/2)2)Q((x-1/2)^2) separates the action on root pairs from the quadratic extensions within the pairs. The reciprocal-polynomial article explains the same construction with opposite pairs. For the standard Euler family, the repeated factor of E5E_5 must first be removed.

References 2

Higher-order Bernoulli polynomials

The order is a separate parameter:

(tet−1)rext=∑n≥0Bn(r)(x)tnn!.\left(\frac{t}{e^t-1}\right)^r e^{xt}=\sum_{n\ge0}B_n^{(r)}(x)\frac{t^n}{n!}.

Adelberg conjectures that the diagonal member Bn(n)B_n^{(n)} is irreducible over Q\mathbb Q for every even n≥2n\ge2, and that Bn(n)(x)/(x−n/2)B_n^{(n)}(x)/(x-n/2) is irreducible for every odd n≥3n\ge3. The linear factor follows from reflection about n/2n/2. This conjecture concerns order nn, not the ordinary family of order one.

Adelberg–Filaseta prove that Eisenstein’s criterion applies to a set of indices with lower density greater than 1/51/5. Filaseta–Luckner later improve this lower bound to (log⁡2)/2(\log 2)/2, and prove the same bound for diagonal higher-order Euler polynomials, defined by replacing t/(et−1)t/(e^t-1) with 2/(et+1)2/(e^t+1). These density results leave the universal Bernoulli assertion unresolved in the cited sources.

References 3

Fixed order

Keeping the order fixed gives a different problem from the diagonal Bn(n)B_n^{(n)}. McCarthy proves that, for each positive integer kk,

B2t(k)(x) is irreducible over Qfor every t≥k.B_{2^t}^{(k)}(x)\text{ is irreducible over }\mathbb Q\quad\text{for every }t\ge k.

More generally, for a fixed prime pp, fixed order kk, and 1≤m<p1\le m<p, the polynomials Bm(p−1)pt(k)B_{m(p-1)p^t}^{(k)} are irreducible once tt exceeds an effective bound. The proof compares coefficients modulo pp as the order decreases and applies Eisenstein’s criterion to a reciprocal polynomial.

Arbitrary degree and order cannot be combined into an unrestricted irreducibility assertion. For example, B2(12)(x)=x2−12x+35=(x−5)(x−7)B_2^{(12)}(x)=x^2-12x+35=(x-5)(x-7).

References 1
  • McCarthy1962 · Theorems 1–2 and Corollary, p. 566; degree-two criterion on p. 567The full page scan confirms t≥k (not t>k). The historical broad conjecture in the final paragraph is not adopted.Source

Questions

OpenPolynomial families

Is every even Bernoulli polynomial irreducible over Q?

Scope. The standard Bernoulli polynomial B_n(x), for every even n≥2; explicitly described as a conjecture by Patel–Siksek (2017), §3.

Carlitz’s criterion proves irreducibility for infinite sets of indices. The unconditional derangement theorem does not settle the general question. The cited paper states the general assertion as a conjecture.

References 2
Discuss this question →
Research directionPolynomial families

What are the Galois groups of the normalized polynomials?

Scope. A stated parity class in the Bernoulli or Appell Euler family, with an explicit squarefree normalization.

The degree-five Euler example shows that the choice of normalization matters. The group on pairs and the kernel of the corresponding quadratic extensions are separate parts of the problem.

References 2
Discuss this question →
OpenPolynomial families

Are the diagonal higher-order Bernoulli polynomials irreducible after removing the forced factor?

Scope. B_n^(n) for even n≥2; B_n^(n)/(x−n/2) for odd n≥3, over Q.

This is Adelberg’s conjecture. Degree one is excluded from the quotient formulation because it leaves a constant. The ordinary odd Bernoulli normalization Bn(x)/[x(x−1)(x−1/2)]B_n(x)/[x(x-1)(x-1/2)] is a different problem.

References 2
Discuss this question →

Related topics

References

  1. Bernoulli and Euler Polynomials (Chapter 24)

    NIST Digital Library of Mathematical Functions. NIST Digital Library of Mathematical Functions.

    Read the source
  2. Galois Theory

    Cox, David A. Edition 2. John Wiley & Sons, 2012.

    Read the source
  3. Note on irreducibility of the Bernoulli and Euler polynomials

    Carlitz, Leonard. Duke Mathematical Journal, 19, 475–481, 1952.

    Read the source
  4. On the Euler and Bernoulli polynomials

    Brillhart, John. Journal für die reine und angewandte Mathematik, 234, 45–64, 1969.

    Read the source
  5. On universal higher order Bernoulli numbers and polynomials

    Kimura, Noriaki. Report of the Research Institute of Industrial Technology, Nihon University, Report 70, 1–22, 2003.

    Read the source
  6. On multiple zeros of Bernoulli polynomials

    Dilcher, Karl. Acta Arithmetica, 134(2), 149–155, 2008.

    Read the source
  7. On powers that are sums of consecutive like powers

    Patel, Vandita and Siksek, Samir. Research in Number Theory, 3, Article 2, 2017.

    Read the source
  8. On mth order Bernoulli polynomials of degree m that are Eisenstein

    Adelberg, Arnold and Filaseta, Michael. Colloquium Mathematicum, 93(1), 21–26, 2002.

    Read the source
  9. Higher Order Bernoulli Polynomials and Newton Polygons

    Adelberg, Arnold. Applications of Fibonacci Numbers, Volume 7, Kluwer Academic Publishers, 1–8, 1998.

    Read the source
  10. On nth order Euler polynomials of degree n that are Eisenstein

    Filaseta, Michael and Luckner, Thomas. Indagationes Mathematicae, 35(1), 76–86, 2024.

    Read the source
  11. Irreducibility of Bernoulli polynomials of higher order

    McCarthy, P. J. Canadian Journal of Mathematics, 14, 565–567, 1962.

    Read the source