\documentclass[11pt]{article} \usepackage[margin=2.6cm]{geometry} \usepackage{amsmath,amssymb,amsthm} \usepackage[T1]{fontenc} \usepackage{lmodern} \usepackage{booktabs} \usepackage[hidelinks]{hyperref} \newtheorem{theorem}{Theorem} \newtheorem{proposition}[theorem]{Proposition} \newtheorem{lemma}[theorem]{Lemma} \theoremstyle{remark} \newtheorem{remark}[theorem]{Remark} \newcommand{\Z}{\mathbb{Z}} \DeclareMathOperator{\ord}{v} \title{The 2-adic valuation of the symmetric\\ Prouhet--Tarry--Escott constant at sizes 13 and 15} \author{David Niedbala Giraudin\\ ORCID 0009-0009-1526-1178} \date{17 November 2026} \begin{document} \maketitle \begin{abstract} For symmetric ideal solutions of the Prouhet--Tarry--Escott problem of odd size $n$ over the $2$-adic integers, the associated constant $C$ was known to satisfy $\ord_2(C)\ge 12$ at $n=13$, with a best known solution at $\ord_2(C)=17$, and $\ord_2(C)\ge 14$ at $n=15$, with no solution recorded. We show that the minimum is exactly $13$ at $n=13$ and exactly $14$ at $n=15$. The lower bounds are computer-assisted: an exhaustive $2$-adic lifting, driven by the row law taken as an exact equality (units included) together with the vanishing of the odd power sums, eliminates every valuation profile below these values. The upper bounds are explicit $2$-adic solutions, certified by the multivariate Hensel lemma. The same procedure reproduces exactly the known values $4,6,9,11$ at $n=5,7,9,11$. A self-contained verifier accompanies the note. \end{abstract} \section{Setting and result} Let $n\ge 3$ be odd. A \emph{symmetric ideal solution of size $n$ over $\Z_2$} is a multiset $A=\{a_1,\dots,a_n\}\subset\Z_2$ such that \begin{equation}\label{eq:ps} p_{2k+1}(A)=\sum_{i=1}^n a_i^{2k+1}=0\qquad(0\le k\le (n-3)/2), \end{equation} and $a_i+a_j\neq 0$ for all $i,j$. Then $A$ and $-A$ are disjoint and have equal power sums of every degree $1,\dots,n-1$, so they form an ideal solution of the Prouhet--Tarry--Escott (PTE) problem. Writing $F(X)=\prod_i(X-a_i)$, condition \eqref{eq:ps} says that $F(X)=XH(X^2)-c$ with $H$ monic, and $F(0)=-c$ gives $c=\prod_i a_i$. The associated constant is $C=\prod_i(X-a_i)-\prod_i(X+a_i)=-2c$, hence \begin{equation}\label{eq:vC} \ord_2(C)=1+\sum_i \ord_2(a_i)=1+t,\qquad t:=\sum_{a_i \text{ even}}\ord_2(a_i). \end{equation} The \emph{profile} of $A$ is the multiset of the valuations of its even entries; $u$ denotes their number and $w=n-u$ the number of odd entries. Since $p_1(A)=0$, $w$ is even and $u$ is odd. We use the conventions of \cite{G57}, where it is shown that $\ord_2(C)\ge 12$ at $n=13$ and an explicit $2$-adic solution with $\ord_2(C)=17$ is given (Kleiman's theorem \cite{Kl} gives $10$); the bound $\ord_2(C)\ge 14$ at $n=15$ is proved in \cite{G58}, which also isolates the two profiles $(7,1,1,1,1)$ and $(5,2,2,1,1)$, $w=8$, as the only open cases at $n=13$, $t=11$. \begin{theorem}\label{thm:main} For symmetric ideal solutions of size $13$ over $\Z_2$, the minimum of $\ord_2(C)$ is $13$. More precisely: \begin{enumerate} \item[(i)] every such solution has $\ord_2(C)\ge 13$; \item[(ii)] there is a solution with $\ord_2(C)=13$, of profile $(4,2,2,1,1,1,1)$, $w=6$. \end{enumerate} In particular $2^{13}$ divides the constant of every symmetric ideal solution of size $13$ over $\Z$. \end{theorem} \begin{theorem}\label{thm:15} For symmetric ideal solutions of size $15$ over $\Z_2$, the minimum of $\ord_2(C)$ is $14$: the lower bound is that of \cite{G58}, recovered independently below, and it is attained by a solution of profile $(5,2,2,1,1,1,1)$, $w=8$. \end{theorem} The lower bounds are computer-assisted (Section~\ref{sec:lower}); the upper bounds are explicit certificates (Section~\ref{sec:upper}). \section{The row law as an exact equality} The following is Proposition~2.1 of \cite{G57} (the evaluation goes back to Rees--Smyth \cite{RS} and to Coppersmith, Mossinghoff, Scheinerman and VanderKam \cite{CMSV}); in \cite{G57} it is used at the level of valuations. Here it is used as an equality, units included. \begin{lemma}[row law]\label{lem:row} Let $A=\{a_1,\dots,a_n\}$ be a symmetric ideal solution of odd size $n$ over $\Z_2$. Then for every $j$ \begin{equation}\label{eq:row} R_j:=\prod_{i=1}^n (a_j+a_i)=2\prod_{i=1}^n a_i . \end{equation} \end{lemma} \begin{proof} $\prod_i(X+a_i)=-F(-X)=XH(X^2)+c$ and $\prod_i(X-a_i)=XH(X^2)-c$, so their difference is the constant $2c$; at $X=a_j$ the second product vanishes, and $c=\prod_i a_i$. \end{proof} \section{The lower bound}\label{sec:lower} Fix $n$, a profile $(f_1,\dots,f_u)$ with $t=\sum f_k$, and suppose that a solution $A$ with this profile exists. Multiplying $A$ by a $2$-adic unit preserves \eqref{eq:ps}, the profile and $\ord_2(C)$; we may therefore assume that one odd entry equals $1$. For $L\ge1$ let $\tilde a_i\in[0,2^L)$ be the residue of $a_i$ modulo $2^L$. \begin{lemma}[necessary conditions at level $L$]\label{lem:nec} The residues $\tilde a_i$ satisfy: \begin{enumerate} \item[(N1)] for an even entry of valuation $f$: $\tilde a_i\equiv 2^f\pmod{2^{f+1}}$ if $L>f$, and $\tilde a_i=0$ if $L\le f$; the odd entries are odd, and the normalised entry equals $1$; \item[(N2)] $p_{2k+1}(\tilde a)\equiv 0\pmod{2^L}$ for $0\le k\le (n-3)/2$; \item[(N3)] for every $j$, with $s_{ji}=(\tilde a_j+\tilde a_i)\bmod 2^L$ and $v'(s)=\ord_2(s)$ if $s\ne0$, $v'(0)=L$: $\sum_i v'(s_{ji})\le 1+t$; if moreover every $s_{ji}\ne0$, then $\sum_i\ord_2(s_{ji})=1+t$, and the units $U_j=\prod_i s_{ji}2^{-\ord_2(s_{ji})}$ satisfy $U_j\equiv U_k \pmod{2^{\min(\pi_j,\pi_k)}}$, where $\pi_j=L-\max_i\ord_2(s_{ji})$. \end{enumerate} \end{lemma} \begin{proof} (N1) is the definition of the profile. (N2): $p_{2k+1}$ has integer coefficients and $p_{2k+1}(A)=0$. (N3): $a_j+a_i\equiv s_{ji}\pmod{2^L}$, so $\ord_2(a_j+a_i)=\ord_2(s_{ji})$ when $s_{ji}\ne0$ and $\ge L$ otherwise, and the unit part of $a_j+a_i$ is congruent to that of $s_{ji}$ modulo $2^{L-\ord_2(s_{ji})}$. By Lemma~\ref{lem:row} and \eqref{eq:vC}, every $R_j$ equals $2\prod a_i$, of valuation $1+t$; the units of the $R_j$ are therefore all equal. \end{proof} All conditions are invariant under permutations of entries of equal valuation (other than the normalised one). Let $S_L$ be the set of tuples of residues modulo $2^L$ satisfying (N1)--(N3), taken up to these permutations. Every element of $S_{L+1}$ reduces modulo $2^L$ to an element of $S_L$, and $S_{L+1}$ is computed exactly by testing all $2^n$ bit extensions of every element of $S_L$ compatible with (N1). \emph{If $S_L=\emptyset$ for some $L$, no solution with that profile exists.} \begin{proposition}\label{prop:sweep} For $n=13$ and each of the $98$ profiles with $u$ odd and $t\le 11$, and for $n=15$ and each of the $135$ profiles with $u$ odd and $t\le 12$, the set $S_L$ is empty for some $L\le 4$. \end{proposition} This is a finite computation (a few seconds); it is replayed by the verifier. By \eqref{eq:vC}, $\ord_2(C)\le 12$ at $n=13$ means $t\le 11$, and then $u\le t2d$, the multivariate Hensel lemma \cite[Theorem~4.1]{Co} gives an exact zero $a$ of $f$ with $a_i\equiv b_i\pmod{2^{380-59}}$. As $321$ exceeds every valuation involved, $a$ has the valuations of $b$, and $a_i+a_j\ne0$ because $\prod_i(b_j+b_i)\not\equiv 0\pmod{2^{321}}$ for every $j$. By \eqref{eq:vC}, $\ord_2(C)=13$. \end{proof} \begin{proposition}\label{prop:sol15} There is a symmetric ideal solution of size $15$ over $\Z_2$ with entries congruent modulo $2^{48}$ to \[ \begin{array}{llll} 47066783113760, & 247202769948004, & 207640644201852, & 240838204572682,\\ 53708447253522, & 173853332231430, & 118966605088222, & 1,\\ 210029373016823, & 54748232418229, & 10125580704469, & 274115814525415,\\ 129269563341747, & 152037868484161, & 50721618074275, & \end{array} \] whose valuations are $5,2,2,1,1,1,1$ and eight times $0$; hence $\ord_2(C)=1+13=14$. \end{proposition} \begin{proof} As for Proposition~\ref{prop:sol}: the verifier contains $15$ integers with $p_{2k+1}\equiv 0\pmod{2^{480}}$ for $0\le k\le 6$; for a suitable choice of seven of the eight odd entries the Jacobian determinant of $(p_1,\dots,p_{13})$ has valuation $d=76$, and $480>2d$. \end{proof} \section{Status} \begin{itemize} \item \textsc{Proved} (computer-assisted): $\ord_2(C)\ge 13$ for every symmetric ideal solution of size $13$ over $\Z_2$, hence over $\Z$ (Proposition~\ref{prop:sweep}). \item \textsc{Proved} (explicit certificate): the value $13$ is attained over $\Z_2$ (Proposition~\ref{prop:sol}). Together: the minimum is exactly $13$ at $n=13$. \item \textsc{Proved}: at $n=15$ the minimum is exactly $14$ (Propositions~\ref{prop:sweep} and~\ref{prop:sol15}, with the lower bound of \cite{G58}). \item \textsc{Verified}: the procedure reproduces the known values at $n=5,7,9,11$, the bound of \cite{G58} at $n=15$, and keeps alive the profiles of all known witnesses. \item \textsc{Open}: no ideal solution of size $13$ or $15$ over $\Z$ is known; whether these values are attained over $\Z$ is open. \end{itemize} \paragraph{Ancillary file.} \texttt{anc/VERIF\_PTE2ADIC\_N13\_N15.py} (Python 3, standard library only) replays the arithmetic self-test, the row law on integer witnesses, the calibration in both directions, the $98$ eliminations at $n=13$ and the $135$ at $n=15$, and the two Hensel certificates. \paragraph{Credits.} The row law in valuation form, the profile framework and the bound $12$ at $n=13$ are from \cite{G57}, the bound $14$ at $n=15$ from \cite{G58}; the evaluation underlying the row law is that of \cite{RS,CMSV}. The values at $n=5,7,9$ are due to Rees--Smyth \cite{RS}, Caley \cite{Ca} and Filaseta--Markovich \cite{FM}; the witness at $n=9$ is Letac's \cite{Le}, those at $n=11$ are from \cite{CMSV}. The research and the writing were carried out with Claude (Anthropic); the computations are replayed by the ancillary verifier. \begin{thebibliography}{99} \bibitem{Co} K.~Conrad, A multivariable Hensel's lemma, expository note, \url{https://kconrad.math.uconn.edu/blurbs/gradnumthy/multivarhensel.pdf}. \bibitem{CMSV} D.~Coppersmith, M.~J.~Mossinghoff, D.~Scheinerman and J.~M.~VanderKam, Ideal solutions in the Prouhet--Tarry--Escott problem, preprint, arXiv:2304.11254. \bibitem{FM} M.~Filaseta and M.~Markovich, Newton polygons and the Prouhet--Tarry--Escott problem, J. Number Theory 174 (2017), 384--400. \bibitem{Ca} T.~Caley, The Prouhet--Tarry--Escott problem, Ph.D. thesis, University of Waterloo, 2012. \bibitem{G57} D.~Niedbala Giraudin, The 2-adic valuation of the constant for symmetric ideal Prouhet--Tarry--Escott solutions, hexagon:2611.00057 (2026). \bibitem{G58} D.~Niedbala Giraudin, Two improved lower bounds for the 2-adic valuation of the constant of the Prouhet--Tarry--Escott problem, hexagon:2611.00058 (2026). \bibitem{Kl} H.~Kleiman, A note on the Tarry--Escott problem, J. Reine Angew. Math. 278/279 (1975), 48--51. \bibitem{Le} A.~Letac, Gazeta Matematic\u{a} 48 (1942), 68--69. \bibitem{RS} E.~Rees and C.~J.~Smyth, On the constant in the Tarry--Escott problem, Lecture Notes in Math. 1415, Springer, 1990, 196--208. \end{thebibliography} \end{document}