Integrality and leading-coefficient conjecture for colored-triangle coefficient polynomials

About 9 years old · traced to

For each fixed k≥0k\geq 0, let ak(n)a_k(n) be the coefficient of qkq^k in Pn(q)P_n(q), regarded as a polynomial in nn as in the polynomiality conjecture. Integrality and leading-coefficient conjecture. The polynomial k!ak(n)k!a_k(n) has integer coefficients, and its leading coefficient is 5k5^k.

The claim is motivated by the observed formulas for the first coefficients and by a local-defect enumeration argument for the leading term. The source does not provide a proof for all kk, so the conjecture remains open.

References

Primary source

Natasha Blitvic and Leonid Petrov, “Colored interlacing triangles and Genocchi medians”, arXiv:2602.04390 (2026).

Additional references

2 papers in this index state this conjecture (2017–2026). The statement above is taken from the most recent of them; the others are arXiv:1708.07998.

Progress summary

Refreshed
Claimed solved

The original paper leaves the conjecture open, while an unverified reader-written argument now claims to prove it for every coefficient order.

Blitvić and Petrov formulate the conjecture in their February 2026 paper on colored interlacing triangles: for fixed kk, k!ak(n)k!a_k(n) should be integral with leading coefficient 5k5^k. The paper explicitly does not prove the assertion for all kk.

Known results

  • The paper conjectures eventual polynomiality of ak(n)a_k(n) with degree kk for fixed kk.
  • Explicit formulas are given through k=5k=5, with computations reported through n=15n=15.
  • A local-defect enumeration explains the predicted asymptotic ak(n)=5kk!nk+O(nk−1)a_k(n)=\frac{5^k}{k!}n^k+O(n^{k-1}), but its converse is presented only as expected.
  • The first coefficient satisfies a1(2)=1a_1(2)=1 and a1(n)=5(n−2)a_1(n)=5(n-2) for n≥3n\geq 3.

Posted attempt

A reader-written argument claims a complete proof: it proposes a generating-function decomposition into irreducible blocks, bounds the size of fixed-energy blocks, and derives both integrality and leading coefficient 5k5^k. This attempt has not been independently verified.

Current status (as of August 2026): The conjecture is open in the primary source, but a complete proof has been posted and remains unverified.

Sources

Solutions 1

ProofThis solution needs a summarySee full solutionHide full solution

Proof for every coefficient order. Write

Pn(q)=21−nT2(n;q)=∑k≥0ak(n)qk.P_n(q)=2^{1-n}T_2(n;q)=\sum_{k\ge0}a_k(n)q^k.

We prove that, for every k≥0k\ge0, the eventual polynomial k!ak(n)k!a_k(n) belongs to Z[n]\mathbb Z[n] and has leading coefficient 5k5^k.

Quotient the depth-two triangles by the n−1n-1 independent energy-preserving interface involutions of source Corollary 2.3 and Proposition 3.4(iii). The normalized polynomial Pn(q)P_n(q) then counts canonical orbit representatives. Every such representative decomposes uniquely into its ordered direct-sum-irreducible blocks, and the energy is additive across blocks. Hence

F(z,q):=∑n≥0Pn(q)zn=11−C(z,q),(1)F(z,q):=\sum_{n\ge0}P_n(q)z^n =\frac1{1-C(z,q)}, \tag{1}

where CC counts nonempty irreducible canonical triangles.

We first establish that each fixed-energy part of CC is a genuine polynomial in zz. Let bjb_j be the jj-th bottom color, and let AjA_j be the active color set immediately before it. Interlacing gives ∣Aj∣=j|A_j|=j and bj∈Ajb_j\in A_j. Define

ej=#{c∈Aj:c>bj},rj=#(Aj∖[j]),hi=#{j≤i:bj>i},E=∑jej.e_j=\#\{c\in A_j:c>b_j\}, \quad r_j=\#(A_j\setminus[j]), \quad h_i=\#\{j\le i:b_j>i\}, \quad E=\sum_j e_j.

The displacement identity and the active-set bound give

D:=∑ihi=∑j(j−bj)+=∑j(bj−j)+≤E,D:=\sum_i h_i =\sum_j(j-b_j)_+ =\sum_j(b_j-j)_+ \le E,

and

rj≤ej+(bj−j)+,∑jrj≤2E.r_j\le e_j+(b_j-j)_+, \qquad \sum_jr_j\le2E.

Whenever hi=ri=ri+1=0h_i=r_i=r_{i+1}=0, the first ii bottom colors are exactly [i][i], the adjacent active sets are [i][i] and [i+1][i+1], and the canonical interface separates the two ordered direct-sum blocks. Thus every noncut is either counted by some positive hih_i, or is adjacent to some positive rjr_j. Consequently,

#{noncuts}≤D+2∑jrj≤5E.\#\{\text{noncuts}\} \le D+2\sum_jr_j \le5E.

An irreducible size-ss configuration has s−1s-1 noncuts. Therefore s≤5E+1s\le5E+1, and

C(z,q)=z+∑j≥1cj(z)qj,cj(z)∈Z[z],deg⁡cj≤5j+1.(2)C(z,q)=z+\sum_{j\ge1}c_j(z)q^j, \qquad c_j(z)\in\mathbb Z[z], \qquad \deg c_j\le5j+1. \tag{2}

Set Q(z,q)=C(z,q)−zQ(z,q)=C(z,q)-z. Expanding (1) gives

F(z,q)=∑ℓ≥0Q(z,q)ℓ(1−z)ℓ+1.F(z,q) =\sum_{\ell\ge0} \frac{Q(z,q)^\ell}{(1-z)^{\ell+1}}.

Because QQ has no constant term in qq,

[qk]F(z,q)=Nk(z)(1−z)k+1,[q^k]F(z,q) =\frac{N_k(z)}{(1-z)^{k+1}},

where

Nk(z)=∑ℓ=0k(1−z)k−ℓ[qk]Q(z,q)ℓ∈Z[z].(3)N_k(z)= \sum_{\ell=0}^k (1-z)^{k-\ell}[q^k]Q(z,q)^\ell \in\mathbb Z[z]. \tag{3}

Writing Nk(z)=∑sdszsN_k(z)=\sum_s d_s z^s, we obtain for all sufficiently large nn,

ak(n)=∑sds(n−s+kk),a_k(n) = \sum_s d_s\binom{n-s+k}{k},

and therefore

k!ak(n)=∑sds∏j=1k(n−s+j)∈Z[n].(4)k!a_k(n) = \sum_s d_s \prod_{j=1}^k(n-s+j) \in\mathbb Z[n]. \tag{4}

Its leading coefficient is

∑sds=Nk(1).\sum_s d_s=N_k(1).

At z=1z=1, all summands in (3) vanish except ℓ=k\ell=k. Since Q=c1(z)q+O(q2)Q=c_1(z)q+O(q^2),

Nk(1)=c1(1)k.(5)N_k(1)=c_1(1)^k. \tag{5}

Finally, source Proposition 3.5 gives a1(2)=1a_1(2)=1 and a1(n)=5(n−2)a_1(n)=5(n-2) for n≥3n\ge3. Comparing coefficients in (1),

[q]F(z,q)=c1(z)(1−z)2=z2+5z3(1−z)2,[q]F(z,q)=\frac{c_1(z)}{(1-z)^2} =z^2+\frac{5z^3}{(1-z)^2},

so

c1(z)=z2+3z3+z4,c1(1)=5.c_1(z)=z^2+3z^3+z^4, \qquad c_1(1)=5.

Substituting into (5), the leading coefficient of k!ak(n)k!a_k(n) is exactly 5k5^k for every k≥0k\ge0. This proves both clauses of Conjecture 4.3.

Source: Blitvić and Petrov, Colored interlacing triangles and Genocchi medians, Conjecture 4.3, https://arxiv.org/abs/2602.04390.