\section{Global finite-order radial TE control with bounded positive index} \label{cc:section} The following construction turns the angular-momentum term in the actual radial Maxwell equation into a controllable finite ensemble. It uses only finite concentric layers with $1\leq\varepsilon\leq q^2<\varepsilon_+$. Its support may grow. The result is approximate global transfer control for every prescribed finite set of TE orders. It is not a bypass theorem. Think of a radial layer as a two-by-two transfer map for each angular order. The same physical layer acts on every order, so the challenge is to control several maps at once rather than tune a separate device for each one. The identity loop below nearly restores the reference transfer while retaining a small dependence on angular order. Repeating such loops at large radius turns that dependence into a useful shear; combinations of shears then give the finite-order approximation theorem. \subsection{An identity loop with an amplified angular derivative} Use $x=k_0r$, the electric TE state $v=(u,u')^T$, and \begin{equation}\label{cc:matrices} X=\begin{pmatrix}0&1\\0&0\end{pmatrix},\qquad Y=\begin{pmatrix}0&0\\1&0\end{pmatrix},\qquad J=X-Y. \end{equation} For $\nu=l(l+1)$ the exact equation in an index-$n$ layer is \begin{equation}\label{cc:actual-ode} v'=(X-n^2Y+\nu x^{-2}Y)v . \end{equation} The state is continuous at dielectric interfaces. No magnetic or surface parameter is being varied. Fix $10$, and let $0\leq\nu\leq V$. Then \begin{equation}\label{cc:one-loop-error} \left\|U_\nu(r)-I-\frac{\nu}{r^2}B_M\right\| \leq \frac{4\nu\ell_Mb_M}{r^3} +\frac{2\nu^2b_M^2}{r^4}\exp(2\nu b_M/r^2). \end{equation} Also $\|U_\nu(r)\|\leq\exp(2\nu b_M/r^2)$. \end{lemma} \begin{proof} Writing the actual fundamental matrix as $U_0W$ gives \[ W'(s)=\frac{\nu}{(r+s)^2}U_0(s)^{-1}YU_0(s)W(s), \qquad W(0)=I. \] The integrated coefficient norm is at most $\beta=2\nu b_M/r^2$. The Volterra series beyond its linear term is bounded by $e^\beta-1-\beta\leq\beta^2e^\beta/2$. Replacing $(r+s)^{-2}$ in that linear term by $r^{-2}$ costs at most $2\nu\ell_M(b_M+d_M)/r^3\leq4\nu\ell_Mb_M/r^3$. Since $U_0(\ell_M)=I$, these are the endpoint errors. The same series bounds $\|W(\ell_M)\|$ by $e^\beta$. \end{proof} \subsection{A macroscopic angular shear in a thin distant annulus} \begin{theorem}[Simultaneous angular shear from positive-index layers] \label{cc:angular-shear} Fix finitely many orders $l_1,\ldots,l_m\geq1$, put $\nu_j=l_j(l_j+1)$ and $V=\max_j\nu_j$, and fix $t\geq0$. For every sufficiently large starting radius $R$ there are finite layers of index one or $q$ in an annulus of dimensionless thickness $o(R)$ whose actual TE transfers converge simultaneously to \begin{equation}\label{cc:angular-shear-target} \exp(t\nu_jY)=I+t\nu_jY,\qquad 1\leq j\leq m . \end{equation} The material bound is independent of $R$, $t$, and the orders. \end{theorem} \begin{proof} For $t=0$ use no layers. Otherwise fix $a\in(1/2,1)$ and set \begin{equation}\label{cc:scales} M=\lfloor a\log_qR\rfloor,\quad h=b_M/R^2,\quad K=\lfloor t/h\rfloor \end{equation} for sufficiently large $R$. Use $K$ consecutive loops, starting at $r_i=R+i\ell_M$, $0\leq i0$, with inverses obtained by interchanging layer order. The singular-value decomposition expresses every real determinant-one matrix as two proper rotations and such a diagonal. Thus finite bounded-index layers realize any common $G$ exactly in the reference equation. For this fixed sequence, the interaction estimate above bounds the angular correction by $O(R^{-2})$, uniformly over the finite orders. Its length is fixed, proving common realizability. Two realizing families concatenate: the first endpoint divided by $R$ tends to one, so use the second family at that endpoint. Total thickness remains $o(R)$ and transfers converge to the product. For closedness let realizable tuples $G^{(n)}\to G$. Choose increasing thresholds $R_n\to\infty$ so that the family for $G^{(n)}$, for every $R\geq R_n$, has transfer error at most $1/n$ and thickness at most $R/n$. On $R_n\leq R