Power-sum relation for arbitrary homogeneous symmetric polynomials

From papers

Let xm={x1,,xm}{\bf x}^m=\{x_1,\ldots,x_m\} and ym={y1,,ym}{\bf y}^m=\{y_1,\ldots,y_m\}. For an arbitrary homogeneous symmetric polynomial Sn(xm)S_n({\bf x}^m) of degree nn, write

Sn(xm)=kCn,kPn,k(xm),Pn,k(xm)=l=1npl(xm)kl,S_n({\bf x}^m)=\sum_{\bf k}C_{n,\bf k}P_{n,\bf k}({\bf x}^m),\qquad P_{n,\bf k}({\bf x}^m)=\prod_{l=1}^n p_l({\bf x}^m)^{k_l},

where i=1niki=n\sum_{i=1}^n i k_i=n, and let π(xm)=j=1mxj\pi({\bf x}^m)=\prod_{j=1}^m x_j. Define

Vn(xm)=Sn(xm)π(xm),Un(xm,ym)=Vn(xm)i=1myimk1Vn(sim),V_n({\bf x}^m)=\frac{S_n({\bf x}^m)}{\pi({\bf x}^m)},\qquad U_n({\bf x}^m,{\bf y}^m)=V_n({\bf x}^m)-\sum_{i=1}^m y_i^{m-k-1}V_n({\bf s}_i^m),

where sim{\bf s}_i^m has entries si,jm=yixjyjxi+xiδijs^m_{i,j}=y_i x_j-y_j x_i+x_i\delta_{ij}. Power-sum relation. The relation

Un(xm,ym)=0,U_n({\bf x}^m,{\bf y}^m)=0,

holds for 0nm10\leq n\leq m-1. Moreover, at ym=1m{\bf y}^m={\bf 1}^m,

Un(xm,1m)=Ynm(xm),nm,U_n({\bf x}^m,{\bf 1}^m)=Y_{n-m}({\bf x}^m),\qquad n\geq m,

where YnmY_{n-m} is homogeneous symmetric of degree nmn-m and

Ynm(xm)=kCn,kYnm,k(xm).Y_{n-m}({\bf x}^m)=\sum_{\bf k}C_{n,\bf k}Y_{n-m,\bf k}({\bf x}^m).

This packages the proposed relations for arbitrary power-sum expansions; the source gives no resolution status.

Progress summary

Open

An October 2025 preprint proposes the relation, but it remains a conditional conjecture rather than a verified solution.

The problem asks whether the proposed identity Un=0U_n=0 holds for 0nm10\le n\le m-1, together with the stated specialization for larger degrees. A preprint dated October 2025 presents this general framework, but does not establish the identity unconditionally.

October 2025 preprint

The manuscript states the general power-sum formulation as Conjecture 6.1 and says it follows assuming earlier Conjectures 3.1 and 3.2. It therefore records a claimed extension to arbitrary homogeneous symmetric polynomials, not a proof; no counterexample, independent verification, referee report, withdrawal, or retraction was found. Its displayed definition also contains the unbound exponent mk1m-k-1.

Current status (as of August 2026): The relation is proposed in a 2025 preprint but remains unproved because its stated derivation depends on earlier unproved conjectures; no counterexample or verification is recorded.

Sources
Sources & referencesView supporting material

Primary source

Boris Y. Rubinstein, “A New Class of Relations for Homogeneous Symmetric Polynomials”, arXiv:2510.25749 (2025).

Solutions 1

Proof

In the primary source, equations (25) and (27) use the exponent mn1m-n-1; the displayed mk1m-k-1 contains an unbound kk and must be corrected accordingly.

Let SS be any symmetric homogeneous polynomial of degree nn in mm variables. Set

Dij=yixjyjxi,(si)j={xi,j=i,Dij,ji,π(x)=ixi,D_{ij}=y_ix_j-y_jx_i, \qquad (s_i)_j= \begin{cases} x_i,&j=i,\\ D_{ij},&j\ne i, \end{cases} \qquad \pi(x)=\prod_i x_i,

and define

U=S(x)π(x)iyimn1S(si)xijiDij.U= \frac{S(x)}{\pi(x)} - \sum_i \frac{y_i^{m-n-1}S(s_i)} {x_i\prod_{j\ne i}D_{ij}}.

Over the coefficient field generated by the yiy_i, the only possible xx-poles are xix_i and DijD_{ij}. On xi=0x_i=0,

S(si)=yinS(xxi=0),jiDij=yim1jixj,S(s_i)=y_i^nS(x|_{x_i=0}), \qquad \prod_{j\ne i}D_{ij} =y_i^{m-1}\prod_{j\ne i}x_j,

so the residue of the ii-th summand cancels the residue of S(x)/π(x)S(x)/\pi(x).

On Dij=0D_{ij}=0, write c=yj/yic=y_j/y_i. Then

xj=cxi,Djr=cDir,sj=cτij(si),x_j=cx_i, \qquad D_{jr}=cD_{ir}, \qquad s_j=c\,\tau_{ij}(s_i),

and hence S(sj)=cnS(si)S(s_j)=c^nS(s_i). The two corresponding residues have ratio

c(mn1)+n(m1)=1,-c^{(m-n-1)+n-(m-1)}=-1,

so they cancel. Therefore UU is a polynomial in xx, homogeneous of degree nmn-m. In particular,

U=0(n<m).U=0\qquad(n<m).

For nmn\ge m, put d=nm+1d=n-m+1. Multiplication by π(y)d\pi(y)^d removes all explicit yy-denominators. With

H=π(x)i<jDij,C=(m2),H=\pi(x)\prod_{i<j}D_{ij}, \qquad C=\binom m2,

the residue cancellations show that the numerator over the common denominator HH is divisible by every pairwise nonassociated irreducible factor xi,Dijx_i,D_{ij}. Thus

R=π(y)dUQ[x1,,xm,y1,,ym],U=Rπ(y)d.R=\pi(y)^dU\in\mathbb Q[x_1,\ldots,x_m,y_1,\ldots,y_m], \qquad U=\frac{R}{\pi(y)^d}.

Moreover,

degxR=nm.\deg_x R=n-m.

The first term of HRHR has yy-degree at most md+Cmd+C; each remaining term has degree at most

(m1)d+n+C(m1)=md+C.(m-1)d+n+C-(m-1)=md+C.

Since HH is homogeneous of yy-degree CC, it follows that

degyRmd=m(nm+1).\deg_yR\le md=m(n-m+1).

Simultaneously permuting the pairs (xi,yi)(x_i,y_i) leaves UU, π(y)d\pi(y)^d, and therefore RR invariant. Specializing y1==ym=1y_1=\cdots=y_m=1 consequently gives a symmetric homogeneous polynomial

Ynm(x)=R(x,1).Y_{n-m}(x)=R(x,\mathbf1).

Finally, linearity in SS gives the asserted decomposition

S=kCn,kPn,kYnm=kCn,kYnm,k.S=\sum_{\mathbf k}C_{n,\mathbf k}P_{n,\mathbf k} \quad\Longrightarrow\quad Y_{n-m} =\sum_{\mathbf k}C_{n,\mathbf k}Y_{n-m,\mathbf k}.

This proves all four source assertions, including the nonspecialized rational form, the degree bound, and simultaneous-permutation symmetry.

0 endorsements
Shivam Patel ·