Overview

  • How many monic polynomials of bounded height have group different from Sₙ?
  • Are irreducible exceptions fewer than reducible polynomials?
  • Why does the reciprocal exceptional count contain a logarithm?

Monic polynomials

Fix n≥3n\ge3 and consider the monic polynomials

f=xn+a1xn−1+⋯+an,∣ai∣≤H.f=x^n+a_1x^{n-1}+\cdots+a_n,\qquad |a_i|\le H.

There are (2⌊H⌋+1)n(2\lfloor H\rfloor+1)^n choices. Let En(H)E_n(H) count the polynomials with repeated roots, together with the separable polynomials whose Galois group on the nn roots is not SnS_n. Thus reducible polynomials are included. Throughout this section, the degree is fixed as the height grows.

Theorem · Bhargava

For fixed n≥3n\ge3,

En(H)=On(Hn−1).E_n(H)=O_n(H^{n-1}).

The (2⌊H⌋+1)n−1(2\lfloor H\rfloor+1)^{n-1} choices with an=0a_n=0 all have the factor xx and give a lower bound of the same order. The exceptional proportion therefore has order H−1H^{-1}.

References 1
  • Bhargava2025 · Theorem 1Sharp upper bound in the fixed-degree box model; the safe stated range here is n≥3.Source

Quadratics

Example

Consider (x+a)(x+b)(x+a)(x+b) for integers 2≤a≤H/22\le a\le\sqrt H/2 and a≤b≤H/(2a)a\le b\le H/(2a). For sufficiently large HH, the coefficient bounds a+b≤Ha+b\le H and ab≤Hab\le H hold. Distinct unordered pairs give distinct polynomials. Summing over the available bb's for each aa gives ≫Hlog⁡H\gg H\log H reducible quadratics.

Thus the bound O(Hn−1)O(H^{n-1}) for all exceptional polynomials fails at n=2n=2. A count restricted to irreducible polynomials with exceptional transitive group excludes these examples.

Irreducible exceptions

Let In(H)I_n(H) count only the irreducible polynomials in the monic box whose Galois group is not SnS_n. The strong van der Waerden conjecture asks whether In(H)=on(Hn−1)I_n(H)=o_n(H^{n-1}).

Preprint result · Anderson et al. (2026)

In(H)=On(Hn−1−1/440)(n≥40).I_n(H)=O_n(H^{n-1-1/440})\qquad(n\ge40).

The degree is fixed as H→∞H\to\infty. Version 2 proves the required twisted Möbius-correlation estimate in this range. Its stronger correlation conjecture would extend the conclusion to n≥8n\ge8. The argument uses a square sieve, differencing and finite-field estimates.

References 1
  • AndersonEtAl2026StrongVdW · Conjecture 1.1; Theorems 1.2, 1.4–1.5; Conjecture 1.6Version 2; I_n here is the paper’s E_n, to distinguish it from the atlas’s count including reducible polynomials.Source

Nonmonic polynomials

Let Rn(H)R_n(H) count degree-nn polynomials in Z[x]\mathbb Z[x] that are reducible over Q\mathbb Q and whose coefficients have absolute value at most HH. The leading coefficient may be any nonzero integer in the given range. Kuba proves

Rn(H)≍nHn(n≥3),R2(H)≍H2log⁡H.R_n(H)\asymp_n H^n\quad(n\ge3),\qquad R_2(H)\asymp H^2\log H.

The total count is 2⌊H⌋(2⌊H⌋+1)n≍nHn+12\lfloor H\rfloor(2\lfloor H\rfloor+1)^n\asymp_n H^{n+1}, so the reducible proportion has order H−1H^{-1} for n≥3n\ge3. The nonmonic count has one additional power of HH relative to the monic count.

The proof counts factorizations using height inequalities and lattice-point estimates. For n≥3n\ge3, all but On(Hn−1(log⁡H)2)O_n(H^{n-1}(\log H)^2) reducible polynomials have an irreducible factor of degree n−1n-1 and hence a linear cofactor.

References 1
  • Kuba2009 · Theorems 1, 2 and 4; Lemma 1The polynomial count is nonmonic and reducibility is over Q, so a nonunit integer content alone does not count as reducibility.Source

Sieves and number fields

At good primes, a Galois group restricts the possible factorization types. Gallagher’s large sieve bounds the number of coefficient vectors whose reductions satisfy these restrictions at many primes.

Bhargava’s sharper bounds use the structure of the exceptional permutation groups and the discriminants of the number fields generated by roots, together with Fourier and geometric-sieve estimates. The count is over coefficient vectors. Many polynomials can define the same number field, so a count by field discriminant answers a different question.

References 2

Counting fields by polynomial height

For a degree-nn number field KK, define Hmin⁡(K)H_{\min}(K) as the least coefficient height of an integral degree-nn polynomial defining KK, with no monicity condition. Arango-Piñeros et al. state the following preprint asymptotic:

#{K/Q:[K:Q]=n, Hmin⁡(K)<X}∼CnXn+1(n≥3),\#\{K/\mathbb Q:[K:\mathbb Q]=n,\ H_{\min}(K)<X\}\sim C_nX^{n+1}\quad(n\ge3),

where fields are counted up to isomorphism and Cn>0C_n>0. For n=2n=2, the asymptotic is (27/π2)X2(27/\pi^2)X^2. The proof controls when two small defining polynomials give the same field, using the action of rational linear fractional transformations.

Counting by field discriminant is a separate problem; Chow–Dietmann give bounds for prescribed even permutation groups in that setting.

References 2

Prescribed permutation actions

Write B(H;G,Ω)B(H;G,\Omega) for the set of monic separable integer polynomials of degree ∣Ω∣|\Omega|, height at most HH, and root action isomorphic to (G,Ω)(G,\Omega).

Reported preprint result · Ben-Porath (2026)

Let 1≤k≤m1\le k\le m. If G≤SmG\le S_m is kk-transitive and [m]k[m]_k denotes the ordered kk-tuples of distinct elements, then for fixed G,m,kG,m,k and every ε>0\varepsilon>0,

∣B(H;G,[m]k)∣≪G,m,k,εHm+1+ε.|B(H;G,[m]_k)|\ll_{G,m,k,\varepsilon} H^{m+1+\varepsilon}.

For S5S_5 on ordered pairs, the degree is 2020 and the bound is Oε(H6+ε)O_\varepsilon(H^{6+\varepsilon}). The natural action has degree five.

The coefficients of ∏α∈A(x−α)\prod_{\alpha\in A}(x-\alpha) generate the field fixed by the setwise stabilizer of the root subset AA. The proof chooses families of subsets to recover the splitting field while controlling the heights. The paper also treats subset actions, primitive wreath-product actions, and regular actions.

References 1
  • BenPorath2026RootSubsets · Theorem 1.3; Lemma 3.1 and Proposition 3.3The degree is the size of the specified permutation set, not the natural permutation degree m.Source

Reciprocal polynomials

A monic reciprocal polynomial of degree 2m2m has mm free coefficients. If the polynomial is separable, its Galois group lies in Wm=C2≀SmW_m=C_2\wr S_m.

Theorem · Anderson–Bertelli–O’Dorney

For each fixed m≥3m\ge3, among separable monic reciprocal polynomials with coefficients bounded by HH, the number whose group is a proper subgroup of WmW_m is

≍mHm−1log⁡H.\asymp_m H^{m-1}\log H.

For nonmonic reciprocal polynomials, the corresponding count is ≍mHmlog⁡H\asymp_m H^m\log H for every m≥1m\ge1.

The logarithmic factor arises from a square-value condition. Polynomials whose Galois group lies in the index-two subgroup of even permutations give the main exceptional contribution. The corrected source leaves the monic quartic case m=2m=2 open.

References 1

Example

For a monic reciprocal quadratic f=x2+ax+1f=x^2+ax+1, the discriminant is a2−4a^2-4. If this equals an integer square d2d^2, then (a−d)(a+d)=4(a-d)(a+d)=4. Since the two factors have the same parity, the integer solutions have a=±2a=\pm2 and d=0d=0. Every separable polynomial in this family has group S2S_2, so the exceptional count in monic degree two is zero.

In the fixed-degree ranges of the ordinary and reciprocal counting theorems, the maximal group occurs with density one.

Questions

OpenCounting and probability

What is the order of the exceptional count for monic reciprocal quartics?

Scope. Separable monic reciprocal quartics over Z, with coefficient height H tending to infinity.

The corrected source proves the monic bound for half-degree m≥3m\ge3 and leaves m=2m=2 open. The nonmonic bound includes quartics.

References 1
Discuss this question →
OpenCounting and probability

Are irreducible exceptions negligible in every fixed degree?

Scope. Monic integer polynomials of fixed degree 5≤n≤39 and coefficient height H→∞.

Does In(H)=on(Hn−1)I_n(H)=o_n(H^{n-1}) hold throughout this range? Here InI_n excludes reducible polynomials. The cited preprint proves the claim for n≥40n\ge40; the cubic and quartic cases were known earlier. Its twisted Chowla conjecture would cover 8≤n≤398\le n\le39, leaving smaller degrees to other arguments.

References 2
Discuss this question →

Related topics

References

  1. Galois groups of random integer polynomials and van der Waerden’s Conjecture

    Bhargava, Manjul. Annals of Mathematics, 201(2), 339–377, 2025.

    Read the source
  2. The strong form of Van der Waerden’s conjecture via twisted Chowla

    Anderson, Theresa C. and Chow, Sam and Dietmann, Rainer and Hokken, David and Koymans, Peter and Lemke Oliver, Robert J. and Sawin, Will and Shusterman, Mark. arXiv, 2026.

    Read the source
  3. On the distribution of reducible polynomials

    Kuba, Gerald. Mathematica Slovaca, 59(3), 349–356, 2009.

    Read the source
  4. The large sieve and probabilistic Galois theory

    Gallagher, P. X. Analytic Number Theory, American Mathematical Society, Proceedings of Symposia in Pure Mathematics, 24, 91–101, 1973.

    Read the source
  5. Counting number fields of fixed degree by their smallest defining polynomial

    Arango-Piñeros, Santiago and Gundlach, Fabian and Lemke Oliver, Robert J. and McGown, Kevin J. and Sawin, Will and Serrano López, Allechar and Shankar, Arul and Varma, Ila. arXiv, 2026.

    Read the source
  6. Enumerative Galois theory for number fields

    Chow, Sam and Dietmann, Rainer. Mathematika, 2026.

    Read the source
  7. Counting Polynomials via Galois Actions on Root Subsets

    Ben-Porath, Or. 2026.

    Read the source
  8. Galois groups of reciprocal polynomials and the van der Waerden-Bhargava theorem

    Anderson, Theresa C. and Bertelli, Adam and O'Dorney, Evan M. Algebra & Number Theory, 20(3), 603–628, 2026.

    Read the source
  9. Enumerative Galois theory for cubics and quartics

    Chow, Sam and Dietmann, Rainer. Advances in Mathematics, 372, Article 107282, 2020.

    Read the source