\section{Paired electric and magnetic radial control} \label{cd:section} The distant finite-order control construction can be tracked in both polarizations simultaneously. Its limiting TE and TM transfers obey a specific inverse-transpose relation. This supplies controlled amplification of an electric-wall reflection sector and suppression of its opposite sector, within any prescribed finite band. It does not bound unselected partial waves or the Green operator of a perturbed device. This distinction matters because a physical dielectric cannot choose its electric and magnetic layer responses independently. The paired equations below follow both responses through the same stack. The resulting relation is later used to identify the amplified and suppressed parity sectors; it is not yet a statement about the response to a source inserted in the core. \subsection{The same physical layers control a paired transfer graph} Retain $X,Y,J,q,\ell_M,b_M,d_M$ from Section~\ref{cc:section}. For TE use $v_{\rm E}=(u,u')^T$ and for TM use $v_{\rm M}=(u,n^{-2}u')^T$ in an index-$n$ layer. With $\nu=l(l+1)$ their exact dimensionless equations are \begin{equation}\label{cd:generators} v_{\rm E}'=(X-n^2Y+\nu x^{-2}Y)v_{\rm E},\qquad v_{\rm M}'=(n^2X-Y+\nu n^{-2}x^{-2}Y)v_{\rm M}. \end{equation} Both states are continuous across each material interface. \begin{lemma}[Paired angular derivatives of the identity loop] \label{cd:loop} For the high--low, then low--high loop of Lemma~\ref{cc:loop-derivative}, its two reference propagators satisfy $U_{0,{\rm M}}=U_{0,{\rm E}}^{-T}$ at every prefix. Their integrated angular derivatives are \begin{equation}\label{cd:loop-derivatives} B_{{\rm E},M}=\begin{pmatrix}0&-d_M\\b_M&0\end{pmatrix}, \qquad B_{{\rm M},M}=\begin{pmatrix}0&-b_M\\d_M&0\end{pmatrix} =-B_{{\rm E},M}^{T}. \end{equation} For both polarizations the integrated interaction-generator norm is $b_M+d_M\leq2b_M$. Consequently the one-loop error \eqref{cc:one-loop-error} and the repeated-loop error \eqref{cc:global-error} apply simultaneously, with the same constants. The limiting repeated-loop transfers are \begin{equation}\label{cd:shear-pair} \bigl(e^{t\nu Y},e^{-t\nu X}\bigr),\qquad t\geq0. \end{equation} \end{lemma} \begin{proof} The two reference generators are negative transposes: $n^2X-Y=-(X-n^2Y)^T$. Differentiating an inverse transpose proves the prefix identity, including across continuous state interfaces. Write the TE prefix as $U=\left(\begin{smallmatrix}a&b\\c&d\end{smallmatrix}\right)$. The TM prefix is $U^{-T}=\left(\begin{smallmatrix}d&-c\\-b&a\end{smallmatrix}\right)$. Its angular interaction matrix, including the actual factor $n^{-2}$, is therefore \begin{equation}\label{cd:tm-interaction} n^{-2}U_{0,{\rm M}}^{-1}YU_{0,{\rm M}} =n^{-2}\begin{pmatrix}cd&-c^2\\d^2&-cd\end{pmatrix}. \end{equation} Its norm is $(c^2+d^2)/n^2$. For completeness, in the forward half with $0\leq j0$, the finite-band core operator converges to $A P_{\rm e}+A^{-1}P_{\rm m}$ after that rephasing. Convergence of all coherent fields and their derivatives is uniform on each fixed compact core subset. The wall boundary identities themselves hold exactly at every stage, by symmetry. \end{theorem} \begin{proof} Take the paired targets $G_l=\operatorname{diag}(A_l,A_l^{-1})$ and restrict starting radii to $R=2\pi n$. The regular radial state tends to $v_l^0=(\sin(-l\pi/2),\cos(-l\pi/2))^T$ in either polarization, because the inner core is vacuum. For even $l$ this is a signed second coordinate vector; for odd $l$ it is a signed first one. Thus the limiting endpoint multiplier on this real state is $A_l^{-1}$ for TE even and TM odd, and $A_l$ for the opposite two cases. At the outer endpoint $X_R$, the incoming radial coefficient obeys the same Wronskian formula in either polarization. The exterior and core are both vacuum, so their state conventions agree there. The Hankel calculation in Corollary~\ref{cc:core-gains} gives $e^{-iX_R}A_l^{\rm in}\to c_l/2$ when the endpoint state is $c_l v_l^0$. Its physical core multiplier, relative to the free wave with the same physical incoming amplitude, is $(2A_l^{\rm in})^{-1}$. After multiplication by $e^{iX_R}$ it therefore tends to $c_l^{-1}$, proving \eqref{cd:wall-gains}. This relative multiplier is valid for TM as well as TE; their different fixed impedance factors cancel in the free/actual comparison. Lemma~\ref{cd:parities} identifies these gains with the two wall sectors for even $m$. All azimuthal indices within an order have the same radial transfer. Finite-band convergence gives uniform compact field and derivative convergence. Radial symmetry preserves reflection and $C_2$ sectors exactly for every finite device, establishing the final assertion. \end{proof} \begin{remark}[What the suppression does and does not provide] \label{cd:scope} For uniform $A$ the limiting multiplier $D_A=A P_{\rm e}+A^{-1}P_{\rm m}$ preserves the vacuum plane flux form pointwise: \begin{equation}\label{cd:flux-cancel} D_A^*Q_z^0(x)D_A=Q_z^0(x)\qquad(z=0). \end{equation} Indeed both diagonal wall blocks vanish pointwise, and the two real factors cancel in the remaining blocks. Amplifying the electric-wall background alone creates no plane passage. The theorem supplies a controlled finite set of opposite-parity TM and TE input responses. It supplies no uniform estimate for unselected angular orders, a total Green operator, a source inside a new inclusion, or a geometry after a nearby inclusion breaks the reflection symmetry. Such estimates are required before a proposed small-transducer construction can be claimed to satisfy physical positive passage and absolute bypass. \end{remark}