Lecture hall position conjecture for the Euler partition bijection

From papers

Let λ=(λ1,λ2,)\begin{aligned} \lambda=(\lambda_1,\lambda_2,\ldots)\end{aligned} be an odd partition, meaning that all its parts are odd, and suppose that its largest part occurs exactly rr times. Let π=φ(λ)\pi=\varphi(\lambda), where φ\varphi is the bijection under discussion. If πj\pi_j determines the lecture hall position, write

rππj(modmj),mj=πjπj+1,rπ>0.r_\pi\equiv \pi_j\pmod {m_j},\qquad m_j=\pi_j-\pi_{j+1},\qquad r_\pi>0.

Lecture hall position conjecture. If π=φ(λ)\pi=\varphi(\lambda), then

r=rπ.r=r_\pi.

This conjecture is supported by substantial numerical evidence. It asserts that the multiplicity of the largest part of an odd partition is preserved as the corresponding lecture hall position under the bijection φ\varphi; no proof or resolution is given in the source.

Progress summary

Open

The conjecture remains an unproved claim that the bijection preserves the largest-part count as a lecture-hall position.

The conjecture asserts that, under the Euler partition bijection from odd-part partitions to distinct-part partitions, the multiplicity of the largest part equals the corresponding lecture-hall position. A 2026 paper introducing this framework poses the relation as an interesting conjecture, without proving or disproving it.

2026 status

The recent paper records the conjecture and its surrounding bijective framework, but no proof, counterexample, or claimed resolution was found in the retrieved material.

Current status (as of August 2026): The conjecture is open; its numerical support and formulation are recorded, but no public proof or disproof has been identified.

Sources
Sources & referencesView supporting material

Primary source

Andrew Y. Z. Wang and Lei Zhang, “Euler's partition theorem and lecture hall partition theorem”, arXiv:2607.16690 (2026).

Solutions 1

Proof

Proof for every odd partition. Let λ\lambda be a nonempty partition into odd parts, let its largest part occur rr times, and write

π=φ(λ)=(π1>>πk>0),πk+1=0,mj=πjπj+1.\pi=\varphi(\lambda) =(\pi_1>\cdots>\pi_k>0), \qquad \pi_{k+1}=0, \qquad m_j=\pi_j-\pi_{j+1}.

By source Theorem 4.2, its lecture-hall length is

L=h(π)=max1jk(πjmj+j1).L=\ell_h(\pi) = \max_{1\le j\le k} \left( \left\lceil\frac{\pi_j}{m_j}\right\rceil+j-1 \right).

Define the maximizing index set

I={j:πjmj+j1=L}.I= \left\{ j: \left\lceil\frac{\pi_j}{m_j}\right\rceil+j-1=L \right\}.

For jIj\in I, write uniquely

πj=(Lj)mj+rj,1rjmj.\pi_j=(L-j)m_j+r_j, \qquad 1\le r_j\le m_j.

The source defines the lecture-hall position by choosing a maximizing index with the largest rjr_j, resolving any remaining tie by the smallest index. Thus

rπ=maxjIrj.(1)r_\pi=\max_{j\in I}r_j. \tag{1}

Let τ\tau be the source's descent on odd partitions. Its gap-free coding sequence is

(bo(λ),bo(τλ),bo(τ2λ),),\bigl( b_o(\lambda), b_o(\tau\lambda), b_o(\tau^2\lambda), \ldots \bigr),

and φ(λ)\varphi(\lambda) is the conjugate partition of that sequence. Applying τ\tau deletes its first entry, hence removes the first column of the conjugate:

φ(τsλ)=(π1s,,πks)>0=:π(s).(2)\varphi(\tau^s\lambda) = (\pi_1-s,\ldots,\pi_k-s)_{>0} =:\pi^{(s)}. \tag{2}

Source Theorem 7.6 proves, for every nonempty odd partition μ\mu,

g(μ)=2h(φ(μ))1,(3)g(\mu)=2\ell_h(\varphi(\mu))-1, \tag{3}

where g(μ)g(\mu) is its largest part. Each descent replaces exactly one occurrence of the current largest odd part tt by t2t-2, or deletes one part if t=1t=1. Therefore the original largest part remains unchanged for exactly rr successive states. By (2)–(3),

r=min{s1:h(π(s))<L},(4)r= \min\{s\ge1:\ell_h(\pi^{(s)})<L\}, \tag{4}

with h()=0\ell_h(\varnothing)=0.

It remains to compute this first-drop time intrinsically. First LkL\ge k, since the contribution of j=kj=k is kk. Whenever jj remains nonterminal after subtracting ss, its gap is unchanged and its contribution is

Fj(s)=πjsmj+j1.F_j(s) = \left\lceil\frac{\pi_j-s}{m_j}\right\rceil+j-1.

If jIj\in I and j<kj<k, then

πj+1=(Lj1)mj+rjrj.\pi_{j+1} = (L-j-1)m_j+r_j \ge r_j.

Thus for 0s<rj0\le s<r_j, the index remains nonterminal and

Fj(s)=L1+rjsmj=L.F_j(s) = L-1+ \left\lceil\frac{r_j-s}{m_j}\right\rceil =L.

For srjs\ge r_j, its contribution is at most L1L-1 while nonterminal; if it becomes terminal, its contribution is j<Lj<L.

The index j=kj=k can belong to II only when L=kL=k, in which case rk=πkr_k=\pi_k. Its contribution remains LL precisely for s<rks<r_k and disappears afterward. An initially nonmaximizing index never becomes maximizing: its nonterminal contribution is nonincreasing, and a terminal contribution is j<Lj<L, unless j=k=Lj=k=L, which was already a maximizing index.

Consequently,

h(π(s))=Ls<maxjIrj.\ell_h(\pi^{(s)})=L \quad\Longleftrightarrow\quad s<\max_{j\in I}r_j.

Combining this with (1) and (4) gives

r=rπ.\boxed{r=r_\pi.}

This proves Conjecture 8.2 for every odd partition, including partitions consisting entirely of ones.

Source: Wang and Zhang, Euler's partition theorem and lecture hall partition theorem, Conjecture 8.2, https://arxiv.org/abs/2607.16690.

0 endorsements
Shivam Patel ·