\section{Exclusion of a simple quadratic pole of residue rank four} \label{sec:sr4-exclusion} We work in the rank-eight coherent-partner construction of Section~\ref{sec:coherent-transpose}, with $\Lambda$ indefinite as in Section~\ref{sec:assembly}. The coefficient spectral form $H=\Lambda^{-1}$ is a nondegenerate indefinite real form on the entire $\mathbb R^8$. The actual lower pencil $P(Y)$ is a generically invertible linear $8\times8$ matrix for $Y\in X^\perp$, and the correction $\Omega(Y)$ is skew-symmetric with $P\Omega$ polynomial. The residue identities \eqref{eq:cp-residue} give $PW=0$ and $WHW^T=0$ for a leading polar coefficient $W$. The real-plane pole classification, Proposition~\ref{prop:rp-denominator-types}, used as a separate input, says that a genuine irreducible real quadratic pole factor $Q$ is positive semidefinite of rank $\rho\ge2$, after changing its sign if necessary. We do not infer this positivity from the denominator degree bound. \begin{proposition}\label{prop:sr4-exclusion} Let $\nu=8$ with $H$ indefinite, and let $X$ be a regular real anchor at which Proposition~\ref{prop:rp-denominator-types} applies. A simple pole of $\Omega$ along a real irreducible positive semidefinite quadratic $Q$ cannot have skew residue rank four over the fraction field of $\mathbb R[Y]/(Q)$. \end{proposition} \begin{proof} The argument first produces a normalized linear matrix factorization. The actual radial row then contradicts the dimensions available for its radical coefficients. \emph{Degree saturation and a linear companion.} The second bound of Lemma~\ref{lem:cp-dvr-budget} makes $Q^4$ divide $\det P$. Both have degree eight, so $\det P=cQ^4$ for a nonzero real constant $c$. At the prime $(Q)$ the four-dimensional residue image lies in $\ker P$. Thus at least four Smith exponents are positive, and their sum is four. They are exactly four ones and four zeros. Every seven-minor is divisible by $Q^3$. Consequently \[ P^\sharp=\frac{\operatorname{adj}P}{cQ^3} \] is a real linear matrix and satisfies $PP^\sharp=P^\sharp P=QI$. Modulo $Q$ it has rank four and \[ \operatorname{Im}P^\sharp=\ker P=\operatorname{Im}W. \] Skewness makes the residue identity $WHW^T=0$ equivalent to isotropy of this image for $H$. Therefore \begin{equation}\label{eq:sr4-companion-form} (P^\sharp)^THP^\sharp=QH_0 \end{equation} with $H_0$ constant real symmetric. These arguments take place over the real prime $(Q)$; they also apply when a rank-two $Q$ splits into conjugate linear factors over $\mathbb C$. Choose a real parameter $e$ with $Q(e)=1$ and put $P_e=P(e)$. Replace $P,P^\sharp$ by $P_e^{-1}P,P^\sharp P_e$, respectively. This is an output-row change for $P$; it does not change the coefficient form $H$. The two matrices are now the identity at $e$, and evaluation of \eqref{eq:sr4-companion-form} at $e$ gives \begin{equation}\label{eq:sr4-normalized} PP^\sharp=P^\sharp P=QI,\qquad (P^\sharp)^THP^\sharp=QH,\qquad P(e)=P^\sharp(e)=I. \end{equation} Choose real parameter coordinates such that $e$ is the first coordinate vector and \[ Q=x_0^2+y_1^2+\cdots+y_{\rho-1}^2, \] with remaining coordinates $z_j$ in its radical. Comparing coefficients in \eqref{eq:sr4-normalized} gives \begin{align*} P&=x_0I+\sum_{i=1}^{\rho-1}y_iC_i+\sum_jz_jD_j,\\ P^\sharp&=x_0I-\sum_{i=1}^{\rho-1}y_iC_i-\sum_jz_jD_j. \end{align*} Each $C_i$ is $H$-skew, squares to $-I$, and anticommutes with the other $C_i$. Every member $D$ of the real linear space $\mathcal D=\operatorname{span}_{\mathbb R}\{D_j\}$ is $H$-skew, satisfies $D^2=0$, and anticommutes with every $C_i$ and every other member of $\mathcal D$. The radical coefficients are not assumed to vanish from their square-zero identity alone. \emph{The actual radial row.} Before the row change, let $g_X$ be the kernel-frame vector representing the unit anchor $X$. The coherent radial identity \eqref{eq:cp-radial-skew} is \begin{equation}\label{eq:sr4-radial-row} g_X^TP(Y)=-Y^TM_X. \end{equation} The image of this row, as $Y$ ranges over $X^\perp$, is all of $\mathbb R^8$, because $M_X$ is an orthonormal image frame in $X^\perp$. After the constant row change the same assertion holds with the fixed covector $g_X^TP_e$. Hence the span of all coefficient matrices of the normalized $P$ has dimension at least eight. The displayed expansion of $P$ therefore implies \begin{equation}\label{eq:sr4-radical-lower-bound} 8\le\rho+\dim_{\mathbb R}\mathcal D. \end{equation} This condition uses the full radial row of the actual partner, rather than an arbitrary matrix factorization of $Q$. \emph{Restrictions on the signature of $H$.} Average a positive definite metric over the finite group of signed monomials in the $C_i$. The resulting positive metric $g$ is invariant. The $g$-self-adjoint operator representing $H$ commutes with the $C_i$, so its positive and negative eigenspaces are invariant modules. A single complex structure makes their dimensions even. Two anticommuting complex structures give a quaternionic action, making their dimensions multiples of four. With four generators, $C_1C_2C_3$ is an involution commuting with the first three and anticommuting with $C_4$. Its two eigenspaces are quaternionic under $C_1,C_2$ and are exchanged by $C_4$. Each nonzero invariant module then has dimension a multiple of eight. Since $H$ is indefinite in dimension eight, $\rho\ge5$ is impossible. Up to reversing its sign, the only possibilities are \[ \begin{array}{c|c} \text{signature of }H&\rho\\\hline (2,6)&2\\ (4,4)&2,3,4. \end{array} \] In particular, the signatures $(1,7)$, $(3,5)$ and their reversals are already excluded. \emph{Anti-complex skew maps.} Put $J=C_1$ and identify $\mathbb R^8$ with $\mathbb C^4$ using $J$. The form $H$ is the real part of a Hermitian form whose signature is half the real signature. In a complex basis its matrix $G$ may be chosen real diagonal or real hyperbolic. An operator anticommuting with $J$ has the form $D(v)=B\overline v$, and $H$-skewness is \[ GB+B^TG=0. \] Thus $GB$ is a complex alternating matrix. Its rank is even, and since $D$ is complex antilinear, its real rank is twice the complex rank of $B$; so the real rank of $D$ is a multiple of four. On the other hand, $D^2=0$ and $H$-skewness imply that $\operatorname{Im}D$ is totally $H$-isotropic. Its rank is bounded by the real Witt index of $H$. For signature $(2,6)$ this forces $\mathcal D=0$, contrary to \eqref{eq:sr4-radical-lower-bound} with $\rho=2$. \emph{A single complex generator and signature $(4,4)$.} Suppose $\rho=2$ and choose $0\ne D_0\in\mathcal D$, if one exists. It has real rank four, and $L=\operatorname{Im}D_0=\ker D_0$ is a $J$-invariant maximal isotropic space, of complex dimension two. In a hyperbolic complex basis, with an isotropic complement to $L$, we may write \[ G=\begin{bmatrix}0&I_2\\I_2&0\end{bmatrix},\qquad D_0(v)=B_0\overline v,\qquad B_0=\begin{bmatrix}0&J_2\\0&0\end{bmatrix},\qquad J_2=\begin{bmatrix}0&1\\-1&0\end{bmatrix}. \] Indeed, the invertible upper-right block of $B_0$ is a complex alternating $2\times2$ matrix; a change on $L$ and the dual change on its complement normalize it to $J_2$. Every $D\in\mathcal D$ preserves $L$, by $DD_0=-D_0D$. Writing $D(v)=B\overline v$, skewness gives \[ B=\begin{bmatrix}E&F\\0&-E^T\end{bmatrix},\qquad F^T=-F. \] Anticommutation with $D_0$ says $EJ_2=J_2\overline E^{\,T}$. Its solutions are exactly \[ E=\begin{bmatrix}a&ib\\ic&\overline a\end{bmatrix}, \qquad a\in\mathbb C,\quad b,c\in\mathbb R. \] The upper-left block of $D^2=0$ is $E\overline E=0$, whereas \[ E\overline E=(|a|^2+bc)I_2. \] The quadratic form $|a|^2+bc$ has real signature $(3,1)$. The image of the real linear map $D\mapsto E$ is a linear subspace of its null cone, hence has dimension at most one. Its kernel has only the alternating complex $2\times2$ block $F$, of real dimension two. Therefore $\dim\mathcal D\le3$, also trivially if $\mathcal D=0$. This gives $\rho+\dim\mathcal D\le5$, again contradicting \eqref{eq:sr4-radical-lower-bound}. \emph{Two or more complex generators.} It remains to treat $\rho=3,4$ with signature $(4,4)$. The first two generators make the real eight-space a quaternionic two-space. Put $J'=C_1C_2$ and $T=-J'D$ for $D\in\mathcal D$. Then $J'$ is $H$-skew, $J'^2=-I$, and $J'$ anticommutes with $C_1$ and $C_2$; since $D$ also anticommutes with them, $T$ commutes with $C_1,C_2$. The operator $D$ commutes with $J'$, so $T^2=0$; it is $H$-self-adjoint and commutes with $C_1,C_2$. Thus $T$ is quaternionic-linear. The invariant form $H$ is the real part of a quaternionic Hermitian form of signature $(1,1)$. In a quaternionic basis with that form diagonal, every such $T$ has the form \[ T=\begin{bmatrix}a&b\\-\overline b&d\end{bmatrix}, \qquad a,d\in\mathbb R,\quad b\in\mathbb H. \] Its nilpotence gives zero real trace, namely $4(a+d)=0$. Consequently $d=-a$ and \[ T^2=(a^2-|b|^2)I_2. \] The real linear image of $\mathcal D$ lies in the null cone of this quadratic form of signature $(1,4)$ and hence has dimension at most one. The map $D\mapsto-J'D$ is injective, so $\dim\mathcal D\le1$. For $\rho=3$ and $4$, respectively, this gives $\rho+\dim\mathcal D\le4$ and $5$, both contradicting \eqref{eq:sr4-radical-lower-bound}. All possibilities are excluded. \end{proof} \begin{remark}\label{rem:sr4-scope} The proof concerns a simple quadratic pole with full skew residue rank four in the rank-eight setup. It does not exclude simple rank-two residues, double poles, or quartic factors. The invariant form is the nondegenerate $H$ on the whole coefficient space. A restriction of $H$ to a smaller residue support can be degenerate; the anti-complex rank argument and the signature table above must not be applied to such a restriction without separate justification. \end{remark}