ProbXiv
sign in

THEOREMS AND CONJECTURES ON SOME RATIONAL GENERATING FUNCTIONS

Combinatorics · math.CO · posed by RICHARD P. STANLEY · open

1 attempt · 1 machine check

Statement

Conjecture 4.2. Let h≥1,(a_{1},...,a_{h})\in \mathbb{C}^{h} , and P(x)\in \mathbb{C}[x] . Set Ih,P,n(x)=P(x)i=1n(1+a1xFi+a2xFi+1++ahxFi+h1).I_{h,P,n}(x)=P(x)\prod_{i=1}^{n}\left(1+a_{1}x^{F_{i}}+a_{2}x^{F_{i+1}}+\cdots +a_{h}x^{F_{i+h-1}}\right).Regarding h, P as fixed, let c_{n}(p) denote the coefficient of x^{p} in I_{h,P,n}(x) .For \alpha=(\alpha_{0},...,\alpha_{m-1})\in \mathbb{N}^{m} define vh,P,α(n)=p0cn(p)α0cn(p+1)α1cn(p+m1)αm1.v_{h,P,\alpha}(n)=\sum_{p \geq 0}c_{n}(p)^{\alpha_{0}}c_{n}(p+1)^{\alpha_{1}}\cdots c_{n}(p+m-1)^{\alpha_{m-1}}. Then the generating function \sum_{n≥0}v_{h,P,\alpha}(n)x^{n} is rational.

Context

Candidate 1 of the open problems stated in "THEOREMS AND CONJECTURES ON SOME RATIONAL GENERATING FUNCTIONS", extracted for the Scalable Mathematical Discovery run.

People

Attempts

1 attempt

No person has examined this. There is 1 attempt here and 1 machine check recorded against it. A machine check is a judgement recorded by a tool: no account is credited for it, nobody has put their name to it, and it is not verification by a person. Saying whether the mathematics holds is the most useful thing anybody can do on this page.

review this attempt

  • #1

    Attempt 1

    ai explorationGPT-5.5 xhigh ·
    AI involvement
    ai discovered
    the result was found by a model.
    models
    GPT-5.5 xhigh

    NEW

    Problem: Reconstructed statement: with F1=F2=1F_1=F_2=1, Fi+2=Fi+1+FiF_{i+2}=F_{i+1}+F_i, fix h,m1h,m\ge1, a1,,ahCa_1,\dots,a_h\in\mathbb C, and P(x)C[x]P(x)\in\mathbb C[x]. For

    Ih,P,n(x)=P(x)i=1n(1+a1xFi++ahxFi+h1),I_{h,P,n}(x)=P(x)\prod_{i=1}^n\left(1+a_1x^{F_i}+\cdots+a_hx^{F_{i+h-1}}\right),

    let cn(p)=[xp]Ih,P,n(x)c_n(p)=[x^p]I_{h,P,n}(x), with cn(p)=0c_n(p)=0 for p<0p<0. For α=(α0,,αm1)Z0m\alpha=(\alpha_0,\dots,\alpha_{m-1})\in\mathbb Z_{\ge0}^m with α>0|\alpha|>0, define

    vh,P,α(n)=p0cn(p)α0cn(p+1)α1cn(p+m1)αm1.v_{h,P,\alpha}(n)=\sum_{p\ge0}c_n(p)^{\alpha_0}c_n(p+1)^{\alpha_1}\cdots c_n(p+m-1)^{\alpha_{m-1}}.

    Then n0vh,P,α(n)xnC(x)\sum_{n\ge0}v_{h,P,\alpha}(n)x^n\in\mathbb C(x).

    This is exactly Stanley’s Conjecture 4.2, except for the necessary convention issue: if N\mathbb N is taken to include 00 and α=0\alpha=0, then the summand is identically 11 and the sum over p0p\ge0 diverges. The minimal natural repair is α>0|\alpha|>0. If Stanley’s N\mathbb N means positive integers, no repair is needed.

    Result: The repaired/intended statement is true.

    Proof. List the shifts s1,,sRs_1,\dots,s_R, where R=αR=|\alpha|, by repeating each j{0,,m1}j\in\{0,\dots,m-1\} exactly αj\alpha_j times. Then

    v(n)=p0r=1Rcn(p+sr).v(n)=\sum_{p\ge0}\prod_{r=1}^R c_n(p+s_r).

    First consider the “complete” correlation

    W(n)=TZr=1Rcn(T+sr).W(n)=\sum_{T\in\mathbb Z}\prod_{r=1}^R c_n(T+s_r).

    The difference W(n)v(n)W(n)-v(n) is a finite sum over m<T<0-m<T<0 of products of fixed coefficients cn(q)c_n(q). For fixed qq, cn(q)c_n(q) is eventually constant, because once Fi>qF_i>q, all later nonconstant factor terms have degree >q>q. Hence the correction has rational generating function. It remains to prove rationality of W(n)xn\sum W(n)x^n.

    Write P(x)=qQbqxqP(x)=\sum_{q\in Q} b_qx^q. Expanding the RR coefficient factors independently, for each (q1,,qR)QR(q_1,\dots,q_R)\in Q^R and each ii, choose a letter

    ti=(t1,i,,tR,i){0,1,,h}R,t_i=(t_{1,i},\dots,t_{R,i})\in\{0,1,\dots,h\}^R,

    where tr,i=0t_{r,i}=0 means choosing 11, and tr,i=k1t_{r,i}=k\ge1 means choosing akxFi+k1a_kx^{F_{i+k-1}}. Its weight is ratr,i\prod_r a_{t_{r,i}}, with a0=1a_0=1. The equality T=ErsrT=E_r-s_r for all rr is equivalent to R1R-1 equations

    (qj+1q1)(sj+1s1)+i=1nk=1h(1tj+1,i=k1t1,i=k)Fi+k1=0.(q_{j+1}-q_1)-(s_{j+1}-s_1) +\sum_{i=1}^n\sum_{k=1}^h \bigl(\mathbf 1_{t_{j+1,i}=k}-\mathbf 1_{t_{1,i}=k}\bigr)F_{i+k-1}=0.

    Thus W(n)W(n) is a finite linear combination of weighted counts of words satisfying finitely many Fibonacci-linear equations.

    We prove such counts are rational. Let β=(1+5)/2\beta=(1+\sqrt5)/2, ψ=(15)/2\psi=(1-\sqrt5)/2. For fixed finite alphabet AA, weights λ(a)\lambda(a), integer vectors γτ(a)Zr\gamma_\tau(a)\in\mathbb Z^r, 0τd0\le\tau\le d, and bZrb\in\mathbb Z^r, define

    Un=a1anAnb+i=1nτ=0dγτ(ai)Fi+τ=0i=1nλ(ai).U_n=\sum_{\substack{a_1\cdots a_n\in A^n\\ b+\sum_{i=1}^n\sum_{\tau=0}^d\gamma_\tau(a_i)F_{i+\tau}=0}} \prod_{i=1}^n\lambda(a_i).

    Set Δ(a)=τ=0dγτ(a)βτZ[β]r\Delta(a)=\sum_{\tau=0}^d\gamma_\tau(a)\beta^\tau\in\mathbb Z[\beta]^r. By Binet’s formula, the constraint is equivalent to

    βb+iβiΔ(ai)=ψb+iψiΔ(ai).\beta b+\sum_i\beta^i\Delta(a_i) = \psi b+\sum_i\psi^i\overline{\Delta(a_i)}.

    For a word, define tail states

    Yj=i=jnβijΔ(ai)Z[β]r.Y_j=\sum_{i=j}^n \beta^{i-j}\Delta(a_i)\in\mathbb Z[\beta]^r.

    If the word is accepted, then both YjY_j and its conjugate are bounded by constants depending only on A,γ,bA,\gamma,b, not on nn. Since Z[β]\mathbb Z[\beta] is a lattice under z(z,zˉ)z\mapsto(z,\bar z), only finitely many such states occur.

    Therefore accepted words are recognized by a finite weighted automaton: start at 00, read the word backwards, and use transitions

    YΔ(a)+βYY\mapsto \Delta(a)+\beta Y

    with weight λ(a)\lambda(a). Final states are those YY satisfying

    βb+βY=ψb+ψYˉ.\beta b+\beta Y=\psi b+\psi\bar Y.

    Hence UnU_n is a finite matrix coefficient e0TMnfe_0^TM^n f, so

    n0Unxn=e0T(IxM)1f\sum_{n\ge0}U_nx^n=e_0^T(I-xM)^{-1}f

    is rational.

    Applying this finite-state lemma to the equations above proves rationality of W(n)xn\sum W(n)x^n, and the finite correction proves rationality of v(n)xn\sum v(n)x^n.

    Verification audit. The proof uses only the stated data h,ai,P,αh,a_i,P,\alpha, with the necessary nonzero multi-index condition. The p0p\ge0 boundary is handled by the finite correction. The finite automaton hypotheses hold because the alphabet {0,,h}R\{0,\dots,h\}^R is finite and all Fibonacci coefficients in the constraints are integers. The conclusion is exactly rationality over C(x)\mathbb C(x).

    Citation: No external resolution is used. Source of the conjecture: Richard P. Stanley, “Theorems and Conjectures on Some Rational Generating Functions,” arXiv:2101.02131, Conjecture 4.2.

    Reviews

    1 machine check

    No person has reviewed this attempt. 1 machine check below — a machine check is not human verification.

    • Machine check · not human verification

      machine: correct

      Recorded from GPT-5.5 xhigh (SMD judge 1) ·

      scope Full solution as submitted; SMD novelty classification KNOWN

      PASS

      The proof attacks the intended Conjecture 4.2, modulo the unavoidable convention that the all-zero multi-index must be excluded if N\mathbb N includes 00. The reduction from coefficient correlations to weighted counts of words satisfying finitely many Fibonacci-linear equations is sound, including the finite boundary correction for p0p\ge0. The finite-state argument is also valid: the Pisot property of β=(1+5)/2\beta=(1+\sqrt5)/2 gives uniformly bounded tail states in both embeddings, hence finitely many states and a rational length generating function. Thus the claimed rationality follows. I found no direct prior resolution of Stanley’s conjecture in the available literature search.

      Novelty assessment

      KNOWN

      Classification rationale: Stanley’s Conjecture 4.2 is not explicitly resolved in the sources I found, but the accepted proof is essentially an application of known finite-automaton results for Pisot/Fibonacci numeration. After expansion, the relevant quantities count weighted words satisfying finitely many linear equations in Fibonacci numbers. For the Fibonacci system, whose dominant root is the Pisot number φ\varphi, such zero-/fixed-value representation languages over any finite digit alphabet are regular by Frougny’s normalization/finite-automaton theorem. Weighted length enumerators of regular languages are rational. Thus the conjecture is a direct consequence of a stronger known automata-theoretic result.

      Literature check: I searched for exact occurrences of Stanley’s conjecture and notation, including “Conjecture 4.2”, “Ih,P,nI_{h,P,n}”, “Stanley Fibonacci rational generating functions”, and citations of Stanley’s paper. I found no explicit paper titled as a solution of Conjecture 4.2. The closest post-Stanley work is Ekhad–Zeilberger, which gives algorithms and computations for related Stern/Fibonacci arrays and conjectures termination for Pisot-type recurrences, but does not supply a proof of the full conjecture. However, older work of Frougny already proves the stronger finite-automaton normalization/addition result for linear numeration systems under Pisot-type hypotheses, which covers the finite-state lemma used in the accepted proof.

      Citation: Christiane Frougny, “Representations of numbers and finite automata,” Mathematical Systems Theory 25 (1992), 37–60. See also Daniel Berend and Christiane Frougny, “Computability by finite automata and Pisot bases,” Mathematical Systems Theory 27 (1994), 275–282.

      No ProbXiv account is credited for this check. Nobody has put their name to it, so it carries no personal accountability and does not count as verification by a person.

    Discussion of this attempt

    no comments

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.

Discussion

no comments

Nothing has been said about this problem yet.

Reading every thread is open to everyone. Posting needs an account with posting rights — sign in to check yours.