\section{Orbit averaging and a sieve criterion}\label{sec:orbit-criterion} The two elementary estimates below isolate the averaging and small/large-prime split used in the local sieves. \begin{proposition}\label{prop:orbit-averaging} Let a finite group $G$ act on the right on a finite set $Y$ equipped with a $G$-invariant probability measure $\pi$. Let $K\le G$, and let $\Omega=y_0K$ be a $K$-orbit of positive measure. On $L^2(Y,\pi)$ put $f=1_\Omega$ and $\mu=\pi(\Omega)$. Suppose $T$ is self-adjoint, commutes with the $G$-action, and satisfies $T1=d1$ with $d>0$. Let $E\subset L^2(Y,\pi)$ be a subspace such that \[ \|Tv\|\le\Lambda\|v\|\quad(v\in E^\perp),\qquad f-\mu1\in E^\perp, \] where $\Lambda\ge0$. Then \[ \left|\frac{(Tf)(y_0)}d-\mu\right| \le\frac\Lambda d(1-\mu). \] In particular, the same right side bounds the upper deviation. \end{proposition} \begin{proof} Right-\(K\) invariance makes \(Tf\) constant on \(\Omega\). Writing \(f=\mu1+h\), one has \(h\perp1\), \(\|h\|_\pi^2=\mu(1-\mu)\), and \[ (Tf)(y_0)=\frac{\langle f,Tf\rangle_\pi}{\mu} =d\mu+\frac{\langle h,Th\rangle_\pi}{\mu}. \] Self-adjointness and \(T1=d1\) remove the cross terms; the spectral hypothesis bounds the last numerator in absolute value by \(\Lambda\mu(1-\mu)\). \end{proof} Only transitivity on \(\Omega\) is used. At \(-11\), its measure is \(\mu=(3/5)\mu_p\); Section~\ref{sec:minus11} checks the norm-character orthogonality and the separate ramified contribution. \begin{proposition}\label{prop:arithmetic-criterion} For each admissible positive integer $n\ge n_0$, let $\mathcal R_n$ be a finite set of $R(n)>0$ candidate representations. Let $E_n$ be an exceptional bad subset, and $B_{p,n}$ bad subsets indexed by a fixed set $\mathcal S$ of primes. Every representation outside $E_n\cup\bigcup_{p\in\mathcal S}B_{p,n}$ is assumed to be good. Suppose, uniformly in the admissible targets and primes, that \[ \frac{|E_n|}{R(n)}\le\rho+\eta(n),\qquad \frac{|B_{p,n}|}{R(n)}\le\mu_p+C_1n^{\delta-1/2},\qquad \frac{|B_{p,n}|}{R(n)}\le C_2n^\delta p^{-2}, \] where $C_1,C_2,\rho,\mu_p,\eta(n)\ge0$, $\delta<1/4$, $\eta(n)\to0$, and \[ \sum_{p\in\mathcal S}\mu_p\le M<\infty,\qquad \rho+M<1. \] Assume also, for every real $P\ge P_0>0$, that \[ \#\{p\in\mathcal S:p\le P\}\le\alpha P+\beta,\qquad \sum_{\substack{p\in\mathcal S\\p>P}}p^{-2} \le\frac\gamma P+\frac\zeta{P^2}, \] with nonnegative constants. For fixed $t>0$, put \[ L_1=\alpha tC_1+\frac{\gamma C_2}{t},\qquad L_2=\beta C_1+\frac{\zeta C_2}{t^2}. \] Every admissible $n\ge\max\{n_0,(P_0/t)^4\}$ satisfying \[ \eta(n)+L_1n^{\delta-1/4}+L_2n^{\delta-1/2}<1-\rho-M \] has a good representation. In particular, every sufficiently large admissible target does. If the constants are effectively known and $\eta$ has an effective modulus $N_\eta(u)$, meaning $\eta(n)\le u$ for all admissible $n\ge N_\eta(u)$, put $\Delta=1-\rho-M$, $a=1/4-\delta$ and $b=1/2-\delta$. An explicit sufficient cutoff is \[ \max\left\{n_0,(P_0/t)^4,N_\eta(\Delta/4), (4L_1/\Delta)^{1/a},(4L_2/\Delta)^{1/b}\right\}. \] \end{proposition} \begin{proof} Split at \(P=t n^{1/4}\), using the first local bound below \(P\) and the second above it. The convergent union bound is \[ \frac{|E_n|+\sum_p|B_{p,n}|}{R(n)} \le\rho+M+\eta(n)+L_1n^{\delta-1/4}+L_2n^{\delta-1/2}. \] A value below one gives a good representation. Both powers tend to zero, and at the effective cutoff each of the three error terms is at most \(\Delta/4\). \end{proof} The prime-count hypotheses always admit the crude choices $\alpha=\gamma=\zeta=1$, $\beta=0$, $P_0=1$, by comparison with integers and a decreasing integral. Arithmetic residue information improves these constants. When \(\alpha C_1,\gamma C_2>0\), the choice \(t=\sqrt{\gamma C_2/(\alpha C_1)}\) minimizes \(L_1\), though not necessarily the complete cutoff. The mixed Eisenstein application uses $\delta=1/5$, $\rho=1/3$, $\eta(n)=2/(3\sqrt n)$, $C_1=9/4$, $C_2=15/2$, $M=3/20$ and $(\alpha,\beta,\gamma,\zeta)=(2/15,17/15,2/15,8/5)$. Section~\ref{sec:infinite} uses $t=2$ and the sharper direct inequality there to obtain $10^8$.