Fixed-perimeter tipping-point conjecture for two residue classes

For positive integers dd, define hd(a)(n)h_d^{(a)}(n) as the number of partitions of nn into parts that are dd-distinct and at least aa, and define m(m1,m2)(n)\ell_m^{(m_1,m_2)}(n) as the number of partitions of nn into parts congruent to m1m_1 or m2m_2 modulo mm. For positive integers 0<m1<m2m0<m_1<m_2\leq m and 0<ad0<a\leq d, the fixed-perimeter tipping-point conjecture.

limn(hd(a)(n)m(m1,m2)(n))={+m>2d+2,m<2d+2.\lim_{n\to\infty}\left(h_d^{(a)}(n)-\ell_m^{(m_1,m_2)}(n)\right)= \begin{cases} +\infty & m>2d+2,\\ -\infty & m<2d+2. \end{cases}

This conjecture proposes a sharp asymptotic tipping point at m=2d+2m=2d+2, motivated by numerical evidence and by known comparisons between fixed-perimeter partition functions. The supplied statement does not assert a result in the boundary case m=2d+2m=2d+2.

Sources & referencesView supporting material

Primary source

Gabriel Gray, Emily Payne, Holly Swisher and Ren Watson, “Fixed perimeter analogues of some partition results”, arXiv:2502.12394 (2025).

Additional references

2 papers in this index state this conjecture (2025). The statement above is taken from the most recent of them; the others are arXiv:2502.11929.

Progress summary

Refreshed
Open

The conjecture remains open: numerical evidence supports the proposed cutoff, but no proof or counterexample has been publicly reported.

The conjecture appears as Conjecture 1.51.5 in a preprint dated February 2025, predicting opposite eventual dominance on the two sides of the threshold m=2d+2m=2d+2. The boundary case is explicitly left more complicated and unresolved apart from selected cases.

Known results

  • Straub, 2016: established the fixed-perimeter identity relating odd-part and distinct-part partitions to Fibonacci numbers.
  • Fu and Tang, 2018: generalized Straub’s identity.
  • Chen et al., 2024: refined these results for corresponding dd-distinct and residue-class counting functions.
  • Kang and Kim, 2021: proved the analogous tipping-point theorem for ordinary fixed-size partition functions.

February 2025 status

The authors present the strict-inequality assertions as conjectural and supported by experimentation, not as proved theorems. Proposition 1.61.6 settles selected boundary comparisons, while other boundary regimes remain unclear; no later source reports a proof, counterexample, or verification.

Current status (as of August 2026): The conjecture remains open; only selected boundary-case comparisons are proved, and neither strict-inequality assertion has a reported proof or counterexample.

Sources

Solutions 1

Proof

Full proof, with exact exponential asymptotics. Let s=d+1s=d+1, u=m1u=m_1, and v=m2v=m_2. The source's proved generating functions (equations (4.4) and (5.2)) give

H(z)=n1hd(a)(n)zn=za1zzs,L(z)=n1m(u,v)(n)zn=zu(1z)+zv(1z)2zm.H(z)=\sum_{n\ge1}h_d^{(a)}(n)z^n=\frac{z^a}{1-z-z^s},\qquad L(z)=\sum_{n\ge1}\ell_m^{(u,v)}(n)z^n =\frac{z^u(1-z)+z^v}{(1-z)^2-z^m}.

For each real t>0t>0, denote by ρt(0,1)\rho_t\in(0,1) the unique solution of 1ρt=ρtt1-\rho_t=\rho_t^t. The map tρtt\mapsto\rho_t is strictly increasing: if t2>t1t_2>t_1, then

1ρt1ρt1t2=ρt1t1ρt1t2>0.1-\rho_{t_1}-\rho_{t_1}^{t_2} =\rho_{t_1}^{t_1}-\rho_{t_1}^{t_2}>0.

The unique dominant pole of HH is ρs\rho_s. Indeed, every zero zz of 1zzs1-z-z^s with zρs|z|\le\rho_s satisfies

1=z+zsz+zsρs+ρss=1,1=|z+z^s|\le |z|+|z|^s\le\rho_s+\rho_s^s=1,

and equality forces z=ρsz=\rho_s. The pole is simple.

Similarly, writing τ=ρm/2\tau=\rho_{m/2}, every zero of (1z)2zm(1-z)^2-z^m with zτ|z|\le\tau satisfies

1=zm(1z)2zm(1z)2τm(1τ)2=1.1=\left|\frac{z^m}{(1-z)^2}\right| \le\frac{|z|^m}{(1-|z|)^2} \le\frac{\tau^m}{(1-\tau)^2}=1.

Equality in 1z1z|1-z|\ge1-|z| forces z=τz=\tau. This pole is simple, and its numerator is strictly positive. Partial fractions therefore give

hd(a)(n)ρsa11+sρss1ρsn,m(u,v)(n)τu(1τ)+τvτ(2(1τ)+mτm1)τn.h_d^{(a)}(n)\sim \frac{\rho_s^{a-1}}{1+s\rho_s^{s-1}}\rho_s^{-n}, \quad \ell_m^{(u,v)}(n)\sim \frac{\tau^u(1-\tau)+\tau^v} {\tau\left(2(1-\tau)+m\tau^{m-1}\right)} \tau^{-n}.

Both leading constants are positive. If m>2sm>2s, strict monotonicity gives ρs<τ\rho_s<\tau, so hd(a)(n)m(u,v)(n)CHρsn+h_d^{(a)}(n)-\ell_m^{(u,v)}(n)\sim C_H\rho_s^{-n}\to+\infty. If m<2sm<2s, then τ<ρs\tau<\rho_s, so the difference is asymptotic to CLτn-C_L\tau^{-n}\to-\infty. Since 2s=2d+22s=2d+2, this proves the conjecture for every permitted d,a,m,u,vd,a,m,u,v.

Additional critical-case information. When m=2sm=2s, the common denominator factors as

(1z)2z2s=(1zzs)(1z+zs).(1-z)^2-z^{2s}=(1-z-z^s)(1-z+z^s).

At ρ=ρs\rho=\rho_s, the numerator of HLH-L equals

J=2ρa+sρu+sρv.J=2\rho^{a+s}-\rho^{u+s}-\rho^v.

Whenever J0J\ne0,

hd(a)(n)2s(u,v)(n)J2ρs+1(1+sρs1)ρn,h_d^{(a)}(n)-\ell_{2s}^{(u,v)}(n) \sim \frac{J}{2\rho^{s+1}(1+s\rho^{s-1})}\rho^{-n},

so its sign is determined explicitly by JJ; the choice (u,v)=(a,a+s)(u,v)=(a,a+s) gives exact equality for every nn.

Source: Gray, Payne, Swisher, and Watson, Fixed perimeter analogues of some partition results, Discrete Mathematics 349 (2026), 114968, Conjecture 1.5.

0 endorsements
Shivam Patel ·