The multipartite orientation conjecture for the multiplicity of 1-1

Let DD be an orientation of a complete multipartite graph, and let rr be the number of its parts having odd size. Write AD(t)A_D(t) for the Eulerian polynomial of DD, and let mult(AD(t),1)\operatorname{mult}(A_D(t),-1) denote the multiplicity of 1-1 as a root of this polynomial. Multipartite orientation conjecture. One has

mult(AD(t),1)=r2.\operatorname{mult}(A_D(t),-1)=\left\lfloor\frac{r}{2}\right\rfloor.

This conjecture would generalize the paper's result that every nn-vertex tournament has 1-1 as a root of multiplicity exactly n/2\lfloor n/2\rfloor. Its status is not resolved in the supplied text.

Progress summary

Solved

A posted argument claims a complete proof of the conjecture, but no independent verification has been found.

Celano, Sieger, and Spiro (2023) conjectured that the multiplicity of 1-1 in the Eulerian polynomial of every orientation of a complete multipartite graph equals r/2\left\lfloor r/2\right\rfloor, where rr is the number of odd-sized parts.

Known results

  • Celano, Sieger, and Spiro (2023): for every tournament on nn vertices, the multiplicity is exactly n/2\left\lfloor n/2\right\rfloor.

Posted attempt

An argument claims a complete proof for every orientation and all part sizes, using a pairing expansion at t=1t=-1 and parity of minimal pairings; it also claims an exact 22-adic valuation for the first nonzero coefficient. The argument has not been independently verified.

Current status (as of August 2026): the tournament case is established, while the multipartite conjecture has a complete-proof claim but no independent verification or corroborating publication.

Sources
Sources & referencesView supporting material

Primary source

Kyle Celano, Nicholas Sieger and Sam Spiro, “Eulerian Polynomials for Digraphs”, arXiv:2309.07240 (2023).

Solutions 1

Proof

Proof for every orientation, with the exact 2-adic valuation of the first nonzero coefficient.

Let G=Ks1,,shG=K_{s_1,\ldots,s_h}, let DD be an arbitrary orientation, and put

n=isi,m=n/2,r=#{i:si is odd},k=r/2.n=\sum_i s_i,\qquad m=\lfloor n/2\rfloor,\qquad r=\#\{i:s_i\text{ is odd}\},\qquad k=\lfloor r/2\rfloor.

Write

AD(t)=πS(V)tdesD(π).A_D(t)=\sum_{\pi\in\mathfrak S(V)} t^{\operatorname{des}_D(\pi)}.

Partition each ordering into mm consecutive unordered two-vertex blocks, followed by a final singleton when nn is odd. For an ordered block configuration PP, let e(P)e(P) count blocks whose vertices lie in different multipartite parts, and let bD(P)b_D(P) count descents on edges between distinct blocks. Summing over the two internal orders of every block gives the exact identity

AD(t)=PtbD(P)2me(P)(1+t)e(P).(1)A_D(t) =\sum_P t^{b_D(P)}2^{m-e(P)}(1+t)^{e(P)}. \tag{1}

Indeed, an internal nonedge contributes 22, while an internal edge contributes 1+t1+t.

Every odd-sized part must provide an unpaired vertex unless it contains the final singleton. Consequently e(P)ke(P)\ge k, and hence (1+t)kAD(t)(1+t)^k\mid A_D(t). Equality e(P)=ke(P)=k holds precisely when every even part is paired internally and every odd part contributes one mixed-block vertex, except that for odd nn one odd part instead contributes the final singleton.

Set ck=[zk]AD(1+z)c_k=[z^k]A_D(-1+z). Formula (1) gives

ck=2mke(P)=k(1)bD(P).(2)c_k =2^{m-k} \sum_{e(P)=k}(-1)^{b_D(P)}. \tag{2}

In a minimal configuration, two distinct mixed blocks use disjoint multipartite parts. For any two blocks B,CB,C, their number of connecting edges is

eG(B,C)=4iBViCVi,e_G(B,C) =4-\sum_i|B\cap V_i|\,|C\cap V_i|,

which is even: if either block is homogeneous, every nonzero summand is even; if both are mixed, their parts are disjoint. Exchanging adjacent blocks therefore changes bD(P)b_D(P) by an even integer. Thus all m!m! orderings of a fixed unordered minimal pairing MM have the same sign, and

ck=2mkm!MMminϵD(M),ϵD(M){1,1}.(3)c_k =2^{m-k}m! \sum_{M\in\mathcal M_{\min}}\epsilon_D(M), \qquad \epsilon_D(M)\in\{-1,1\}. \tag{3}

The number of unordered minimal pairings is

Mmin={(r1)!!si even(si1)!!si oddsi!!,n even,r!!si even(si1)!!si oddsi!!,n odd,|\mathcal M_{\min}| = \begin{cases} (r-1)!! \displaystyle\prod_{s_i\ {\rm even}}(s_i-1)!! \displaystyle\prod_{s_i\ {\rm odd}}s_i!!, & n\ {\rm even},\\[2mm] r!! \displaystyle\prod_{s_i\ {\rm even}}(s_i-1)!! \displaystyle\prod_{s_i\ {\rm odd}}s_i!!, & n\ {\rm odd}, \end{cases}

with (1)!!=1(-1)!!=1. Every factor is odd. Therefore

MMminϵD(M)Mmin1(mod2).\sum_{M\in\mathcal M_{\min}}\epsilon_D(M) \equiv|\mathcal M_{\min}| \equiv1\pmod2.

In particular ck0c_k\ne0, and (3) proves the stronger conclusions

ordt=1AD(t)=r2,v2 ⁣([zk]AD(1+z))=mk+v2(m!).\boxed{ \operatorname{ord}_{t=-1}A_D(t) =\left\lfloor\frac r2\right\rfloor, \qquad v_2\!\left([z^k]A_D(-1+z)\right) =m-k+v_2(m!). }

Both hold for every orientation, all part sizes, and all boundary cases r=0,1r=0,1. The first equality is precisely the conjecture.

Source: Celano, Sieger, and Spiro, Eulerian polynomials for digraphs, Combinatorial Theory 5 (2025), article 15, Conjecture 6.4, https://doi.org/10.5070/C65165026 ; https://arxiv.org/abs/2309.07240 .

0 endorsements
Shivam Patel ·