\section{The Eisenstein proof}\label{sec:eisenstein} \subsection{Arithmetic preliminaries and split-prime cores} Let $Q(x,y)=x^2-xy+y^2$, $\omega^2+\omega+1=0$ and $R=\Z[\omega]$. Thus $Q(x,y)=N(x+y\omega)$. We use the semigroup $\AA$ defined above and write $\NN=P_{-3}$. Every member of $\AA$ is $1\bmod6$. The ramified dilation $3\!\cdot\!\AA=\{3a:a\in\AA\}$ has residue $3\bmod6$. The elementary norm-value identity separates the two binary problems. Its proof uses only prime splitting and unique factorization. \begin{proposition}\label{prop:primitive-classification} The primitive Eisenstein norms are exactly \[ \NN=\AA\sqcup(3\!\cdot\!\AA). \] \end{proposition} \begin{proof} If an inert prime $p$ divides $Q(x,y)$, it divides both coordinates, since $R/pR$ is a field, including when $p=2$. Also $9\mid Q(x,y)$ forces $3\mid x,y$. Indeed, $3\mid Q(x,y)$ gives $x+y\equiv0\pmod3$; writing $y=-x+3t$ gives \[ Q(x,y)=3(x^2-3xt+3t^2). \] Divisibility by $9$ now forces $3\mid x$, and hence $3\mid y$. Thus a primitive norm has no inert prime factor and contains at most one factor $3$. It belongs to the stated union. Conversely, the triangular Eisenstein lattice has covering radius $1/\sqrt3$, so rounding gives a remainder of norm at most $1/3<1$. Thus $R=\Z[\omega]$ is norm Euclidean and is a unique factorization domain. For a rational prime $q\equiv1\pmod3$, choose a nontrivial cube root $t\in\F_q^\times$. The ideal $(q,\omega-t)$ has quotient $\F_q$; its principal generator $\pi_q$ therefore has norm $q$. The two distinct roots $t,t^{-1}$ yield conjugate, nonassociate factors. Thus $q=\pi_q\bar\pi_q$, up to a unit. For $a=\prod q^{e_q}\in\AA$, choose one factor $\pi_q$ at each $q$ and put $z=\prod\pi_q^{e_q}$. Then $N(z)=a$. The element $(1-\omega)z$ has norm $3a$. Neither element is divisible by a rational prime: a split rational prime would require both conjugate factors, an inert prime cannot divide its norm, and $3$ is associated to $(1-\omega)^2$, whereas at most one factor $1-\omega$ was used. Their coordinates in the integral basis $1,\omega$ are therefore coprime. Finally, the two sets are disjoint, since elements of $\AA$ are prime to $3$. \end{proof} \begin{theorem}\label{thm:main} One has \[ \AA+\AA=\{n>0:n\equiv2\pmod6\},\qquad \AA+(3\!\cdot\!\AA)=\{n>0:n\equiv4\pmod6\}. \] \end{theorem} \paragraph{A reduction that helps both problems.} Put $\PP=\{p\text{ prime}:p\equiv2\pmod3\}$. For either admissible residue class, write $n=gn_0$, where $g\in\AA$ contains every prime-power factor of $n$ with prime $1$ modulo $3$. Since $g\equiv1\pmod6$, the target $n_0$ has the same residue as $n$. A decomposition of either type for $n_0$ scales by $g$ to the corresponding decomposition of $n$. We can therefore concentrate on \begin{equation}\label{eq:reduced} n\equiv2\text{ or }4\pmod6,\qquad p\mid n\Longrightarrow p\in\PP. \end{equation} This reduction will matter in the divisor estimates and in the exact calculation at $3$. The finite certificates cover all targets in their respective ranges, so they need no reduction.