Overview

  • Why does reversing one factor preserve PP*?
  • Why is the resulting polynomial still binary?
  • Which exceptional factors must be treated separately?

The binary model

Let

Pn(x)=1+∑j=1n−1ξjxj+xn,Pr⁡(ξj=0)=Pr⁡(ξj=1)=12,P_n(x)=1+\sum_{j=1}^{n-1}\xi_jx^j+x^n,\qquad \Pr(\xi_j=0)=\Pr(\xi_j=1)=\tfrac12,

with independent ξj\xi_j.

Preprint theorem · Kogler, v2, 14 October 2026

Theorem 1.1 states the unconditional asymptotic

Pr⁡(Pn reducible over Q)=2πn+O(n−1).\Pr(P_n\text{ reducible over }\mathbb Q)=\sqrt{\frac{2}{\pi n}}+O(n^{-1}).

The leading contribution comes from the factor x+1x+1. The theorem concerns reducibility in this binary model. The Galois group and the fair-sign model are separate questions.

References 1

The reciprocal product

For a polynomial PP with nonzero constant term, let P∗(x)=xdeg⁡PP(1/x)P^*(x)=x^{\deg P}P(1/x). Suppose a binary polynomial with leading and constant coefficients 11 factors as P=ABP=AB, with A,B∈Z[x]A,B\in\mathbb Z[x] monic. Put Q=AB∗Q=AB^*. Then

QQ∗=PP∗.QQ^*=PP^*.

Then QQ also has coefficients in {0,1}\{0,1\}, with leading and constant coefficients one.

Proof

The identity follows because reversal is multiplicative and involutive on polynomials with nonzero constant term.

Write P=∑pjxjP=\sum p_jx^j and Q=∑qjxjQ=\sum q_jx^j. The central coefficient of the identity gives ∑qj2=∑pj2\sum q_j^2=\sum p_j^2. Reversal preserves evaluation at 11, so Q(1)=P(1)Q(1)=P(1). Since pj2=pjp_j^2=p_j,

∑jqj(qj−1)=0.\sum_jq_j(q_j-1)=0.

Each summand is a nonnegative integer. Thus each qjq_j equals 00 or 11, and its leading and constant coefficients equal 11.

A factorization with neither factor reciprocal gives Q∉{P,P∗}Q\notin\{P,P^*\}. Thus one bounds such pairs P,QP,Q, and treats reciprocal factors separately.

References 3

Example

Let A=1+x+x3A=1+x+x^3 and B=1+x4+x12B=1+x^4+x^{12}. The supports of P=ABP=AB and Q=AB∗Q=AB^* are

supp⁡(P)={0,1,3,4,5,7,12,13,15},supp⁡(Q)={0,1,3,8,9,11,12,13,15}.\begin{aligned}\operatorname{supp}(P)&=\{0,1,3,4,5,7,12,13,15\},\\\operatorname{supp}(Q)&=\{0,1,3,8,9,11,12,13,15\}.\end{aligned}

In each product, the shifted copies of AA have disjoint supports, so the coefficients are binary. The reciprocal products agree. Yet Q≠PQ\ne P and Q≠P∗Q\ne P^*, since the support of P∗P^* is {0,2,3,8,10,11,12,14,15}\{0,2,3,8,10,11,12,14,15\}.

Thus the autocorrelation encoded by PP∗PP^* can agree for binary polynomials that are distinct even up to reversal.

The congruence estimate

Kogler’s Theorem 1.3 bounds the probability of a nontrivial binary twin even when equality of reciprocal products is replaced by congruence modulo 44. Reduction modulo 22 restricts the possible factors, and the congruence imposes further conditions on opposite coefficient pairs. Cyclotomic factors and reciprocal noncyclotomic factors are treated separately.

The factor-reversal identity also applies in deterministic arguments, including those for polynomials with large gaps.

References 2

Connections and evidence

Claimed · preprint

Binary factorization, Fourier ambiguity, and irreducibility

Factor reversal converts two nonreciprocal factors into a second binary signal with identical Fourier magnitude. A collision bound and two separate divisor estimates give the claimed reducibility asymptotic.

Scope. Pn=1+∑j=1n−1ξjXj+XnP_n=1+\sum_{j=1}^{n-1}\xi_jX^j+X^n, with independent fair ξj∈{0,1}\xi_j\in\{0,1\}, over Q\mathbb Q; n→∞n\to\infty. Fourier comparisons use binary polynomials of the same degree with both endpoints equal to one.

Binary factorization, Fourier ambiguity, and irreducibilityFactor reversal converts two nonreciprocal factors into a second binary signal with identical Fourier magnitude. A collision bound and two separate divisor estimates give the claimed reducibility asymptotic. Connection 1: Reducible binary polynomial leads to OR. Every reducible polynomial in this model has a nontrivial binary twin or a nonconstant reciprocal divisor; the latter is cyclotomic or has a noncyclotomic factor. Connection 2: OR is an input to Nontrivial binary twin. This branch records a possible obstruction, not a conclusion implied separately by reducibility. Connection 3: OR is an input to Reciprocal noncyclotomic divisor. This branch records a reciprocal divisor which is not a product of cyclotomic polynomials. Connection 4: OR is an input to Cyclotomic divisor. This branch records a cyclotomic divisor; the three obstruction events need not be disjoint. Connection 5: Nontrivial binary twin is equivalent to Nontrivial Fourier ambiguity. For binary P and Q of the same degree with Q distinct from P and P*, the identity PP*=QQ* is equivalent to equality of their Fourier magnitudes at every angle. Connection 6: Nontrivial Fourier ambiguity is an input to Modulo-4 collision bound. Exact reciprocal-product equality is included in congruence modulo four, whose nontrivial collision probability is bounded in Theorem 1.3. Preprint claim. Connection 7: Reciprocal noncyclotomic divisor is an input to Reciprocal-divisor bound. Proposition 3.2 bounds the probability of a monic reciprocal divisor that is not a product of cyclotomic polynomials. Preprint claim. Connection 8: Cyclotomic divisor is an input to Cyclotomic estimates. The factor X-1 is impossible, higher-degree cyclotomic divisors contribute O(1/n), and X+1 supplies the leading term. Connection 9: Modulo-4 collision bound is an input to AND. Use the collision estimate together with both divisor estimates. Connection 10: Reciprocal-divisor bound is an input to AND. Use the reciprocal-divisor estimate together with the collision and cyclotomic estimates. Connection 11: Cyclotomic estimates is an input to AND. Use both the upper bound for other cyclotomic factors and the exact probability that P_n(-1)=0. Connection 12: AND leads to Reducibility asymptotic. The union bound gives the upper estimate, while divisibility by X+1 for n≥2 gives the matching leading lower estimate. Preprint claim.Reducible binary polynomial — related articleReducible binarypolynomialORORNontrivial binary twin — related articleNontrivial binary twinReciprocal noncyclotomic divisor — related articleReciprocal noncyclotomicdivisorCyclotomic divisor — related articleCyclotomic divisorNontrivial Fourier ambiguity — related articleNontrivial FourierambiguityModulo-4 collision bound — related articleModulo-4 collisionboundReciprocal-divisor bound — related articleReciprocal-divisor boundCyclotomic estimates — related articleCyclotomic estimatesANDANDReducibility asymptotic — related articleReducibility asymptoticConnection 1: Every reducible polynomial in this model has a nontrivial binary twin or a nonconstant reciprocal divisor; the latter is cyclotomic or has a noncyclotomic factor.Exhaustive obstructionsplit [1]Connection 2: This branch records a possible obstruction, not a conclusion implied separately by reducibility.Possible obstruction [2]Connection 3: This branch records a reciprocal divisor which is not a product of cyclotomic polynomials.Possible obstruction [3]Connection 4: This branch records a cyclotomic divisor; the three obstruction events need not be disjoint.Possible obstruction [4]Connection 5: For binary P and Q of the same degree with Q distinct from P and P*, the identity PP*=QQ* is equivalent to equality of their Fourier magnitudes at every angle.Reciprocal-productidentity [5]Connection 6: Exact reciprocal-product equality is included in congruence modulo four, whose nontrivial collision probability is bounded in Theorem 1.3.Bound the ambiguityevent [6]Connection 7: Proposition 3.2 bounds the probability of a monic reciprocal divisor that is not a product of cyclotomic polynomials.Estimate this divisorevent [7]Connection 8: The factor X-1 is impossible, higher-degree cyclotomic divisors contribute O(1/n), and X+1 supplies the leading term.Isolate the factor X+1[8]Connection 9: Use the collision estimate together with both divisor estimates.Joint input [9]Connection 10: Use the reciprocal-divisor estimate together with the collision and cyclotomic estimates.Joint input [10]Connection 11: Use both the upper bound for other cyclotomic factors and the exact probability that P_n(-1)=0.Joint input [11]Connection 12: The union bound gives the upper estimate, while divisibility by X+1 for n≥2 gives the matching leading lower estimate.Theorem 1.1 [12]
Linked nodes open articles; numbered connections open the evidence below.Open full-size diagram

deduction or information flow equivalence preprint claimAND joint hypothesesOR alternatives

Let Pn={1+∑j=1n−1ξjXj+Xn:ξj∈{0,1}}\mathcal P_n=\{1+\sum_{j=1}^{n-1}\xi_jX^j+X^n:\xi_j\in\{0,1\}\}, and choose PnP_n uniformly. Write P∗(X)=Xdeg⁡PP(1/X)P^*(X)=X^{\deg P}P(1/X). If P=ABP=AB with monic nonconstant integer factors, then Q=AB∗Q=AB^* is binary and QQ∗=PP∗QQ^*=PP^*; it differs from both PP and P∗P^* exactly when neither factor is reciprocal. On the unit circle, PP∗(z)=zn∣P(z)∣2PP^*(z)=z^n|P(z)|^2.

Preprint claim, Kogler v2. Theorem 1.3 states

Pr⁡(∃Q∈Pn∖{Pn,Pn∗}:QQ∗≡PnPn∗(mod4))≤8(3/4)⌊(n−1)/4⌋.\Pr\bigl(\exists Q\in\mathcal P_n\setminus\{P_n,P_n^*\}:QQ^*\equiv P_nP_n^*\pmod4\bigr)\le8(3/4)^{\lfloor(n-1)/4\rfloor}.

This also bounds nontrivial exact Fourier-magnitude ambiguity. It does not alone control every reducibility event: reciprocal noncyclotomic divisors require Proposition 3.2, while cyclotomic divisors require separate estimates. Reducibility is contained in the union of the three displayed obstruction events.

The reciprocal-divisor probability is at most Cexp⁡(−cn/(log⁡n)4)C\exp(-cn/(\log n)^4) for sufficiently large nn. Cyclotomic factors of degree at least two have probability O(n−1)O(n^{-1}). Since Pn(1)>0P_n(1)>0, the only remaining linear factor is X+1X+1. Its probability is 2/(πn)+O(n−3/2)\sqrt{2/(\pi n)}+O(n^{-3/2}). Together these estimates give Theorem 1.1:

Pr⁡(Pn reducible over Q)=2πn+O(n−1).\Pr(P_n\text{ reducible over }\mathbb Q)=\sqrt{\frac{2}{\pi n}}+O(n^{-1}).

The conclusion concerns irreducibility in the binary coefficient model. Galois-group conclusions and other coefficient laws require separate theorems.

Inspect all 12 connections and exact references
Evidence for each connection
ConnectionStatement and source
1Reducible binary polynomial → ORDeduction

Every reducible polynomial in this model has a nontrivial binary twin or a nonconstant reciprocal divisor; the latter is cyclotomic or has a noncyclotomic factor.

2OR → Nontrivial binary twinConstruction / input

This branch records a possible obstruction, not a conclusion implied separately by reducibility.

3OR → Reciprocal noncyclotomic divisorConstruction / input

This branch records a reciprocal divisor which is not a product of cyclotomic polynomials.

4OR → Cyclotomic divisorConstruction / input

This branch records a cyclotomic divisor; the three obstruction events need not be disjoint.

5Nontrivial binary twin ↔ Nontrivial Fourier ambiguityEquivalence

For binary P and Q of the same degree with Q distinct from P and P*, the identity PP*=QQ* is equivalent to equality of their Fourier magnitudes at every angle.

  • B2 · Kogler (2026)Lemma 3.1, p. 7; Fourier-magnitude identity, §1, p. 3; arXiv:2609.26771v2Lemma 3.1 restates the classical binary twin argument and gives its proof.Primary source
6Nontrivial Fourier ambiguity → Modulo-4 collision boundConstruction / input · preprint

Exact reciprocal-product equality is included in congruence modulo four, whose nontrivial collision probability is bounded in Theorem 1.3.

7Reciprocal noncyclotomic divisor → Reciprocal-divisor boundConstruction / input · preprint

Proposition 3.2 bounds the probability of a monic reciprocal divisor that is not a product of cyclotomic polynomials.

8Cyclotomic divisor → Cyclotomic estimatesConstruction / input

The factor X-1 is impossible, higher-degree cyclotomic divisors contribute O(1/n), and X+1 supplies the leading term.

9Modulo-4 collision bound → ANDConstruction / input

Use the collision estimate together with both divisor estimates.

10Reciprocal-divisor bound → ANDConstruction / input

Use the reciprocal-divisor estimate together with the collision and cyclotomic estimates.

11Cyclotomic estimates → ANDConstruction / input

Use both the upper bound for other cyclotomic factors and the exact probability that P_n(-1)=0.

12AND → Reducibility asymptoticDeduction · preprint

The union bound gives the upper estimate, while divisibility by X+1 for n≥2 gives the matching leading lower estimate.

Sources and scope checked .

Questions

Research directionCounting and probability

Does the binary model have full symmetric group with probability tending to one?

Scope. Monic binary polynomials over Q with constant term 1 and independent fair internal coefficients.

The cited preprint states a high-probability irreducibility theorem. A full-SnS_n theorem also requires control of the transitive group and the discriminant square class.

References 1
Discuss this question →

Related topics

References

  1. The Odlyzko-Poonen Conjecture on Irreducibility of Random Polynomials

    Kogler, Constantin. 2026.

    Read the source
  2. On the irreducibility of certain trinomials and quadrinomials

    Ljunggren, Wilhelm. Mathematica Scandinavica, 8, 65–70, 1960.

    Read the source
  3. On the factorization of polynomials with small Euclidean norm

    Filaseta, Michael. Number Theory in Progress, Vol. 1, de Gruyter, 143–163, 1999.

    Read the source
  4. Irreducibility of polynomials with a large gap

    Sawin, Will and Shusterman, Mark and Stoll, Michael. Acta Arithmetica, 192(2), 111–139, 2020.

    Read the source
  5. Irreducibility of random polynomials of large degree

    Breuillard, Emmanuel and Varjú, Péter P. Acta Mathematica, 223(2), 195–249, 2019.

    Read the source