\section{Exclusion of a double quadratic pole} \label{sec:dq-exclusion} We work in the indefinite case $\nu=8$ with the notation of Sections~\ref{sec:coherent-transpose} and~\ref{sec:real-pole-types}: at a regular anchor $X$, for $Y\perp X$, \[ M_Y=M_XF(Y)+K_XP(Y),\qquad F=F_0+\Omega H,\qquad \Omega^T=-\Omega, \qquad H=\Lambda^{-1}, \] where $F_0,P$ are linear, $P$ is generically invertible, and $P\Omega$ is polynomial, because the lower--upper Jacobi block $P\Lambda F^T$ is. \begin{proposition}\label{prop:dq-exclusion} There is no nonempty open set of regular anchors $X$ at which the reduced common denominator of $F$ on $X^\perp$ is $Q_X^2$ for an irreducible real positive semidefinite quadratic $Q_X$ (in particular, no branch $\mathcal U_j$ of Lemma~\ref{lem:asm-partition} with nonempty interior is of this type). \end{proposition} \subsection{A degree bound for an entire skew block} \begin{lemma}\label{lem:dq-block-degree} Let $P_0$ be a real polynomial $m\times b$ matrix with linear entries and generic column rank $b$. Let $\Omega_0$ be a skew rational $b\times b$ matrix such that $P_0\Omega_0$ is polynomial. If $\Omega_0$ has pole order $e$ at a real irreducible quadratic $Q$, then every maximal $b$-minor of $P_0$ is divisible by $Q^{2e}$. In particular, $b\ge4e$. \end{lemma} \begin{proof} Over the discrete valuation ring at $Q$ take a Smith form $P_0=U\left[\begin{smallmatrix}\operatorname{diag}(Q^{a_1},\ldots,Q^{a_b})\\0\end{smallmatrix}\right]V$ with $U,V$ invertible over that ring. The congruent skew matrix $\widehat\Omega=V\Omega_0V^T$ still has pole order $e$; choose an off-diagonal entry $\widehat\Omega_{ij}$ of valuation $-e$. Polynomiality of $P_0\Omega_0$ makes $\operatorname{diag}(Q^{a_1},\ldots,Q^{a_b})\,\widehat\Omega$ integral; its rows $i$ and $j$, which contain the entries $\pm\widehat\Omega_{ij}$, force $a_i,a_j\ge e$, so every maximal minor has valuation at least $2e$. Some maximal minor is nonzero, of degree at most $b$, whereas $\deg Q^{2e}=4e$. \end{proof} Under the double-pole hypothesis this lemma, applied to $P,\Omega$, gives by degree eight \begin{equation}\label{eq:dq-determinant} \det P=c_XQ_X^4,\qquad c_X\ne0. \end{equation} Let $g_X$ be the unit vector with $K_Xg_X=X$ (Section~\ref{sec:coherent-transpose}). The radial row $g_X^TP(Y)=-Y^TM_X$ vanishes for $Y$ in the reduced zero eigenspace $E_0(X)=\ker\Jac_X\cap X^\perp$, so \eqref{eq:dq-determinant} gives $Q_X(Y)=0$ there, and semidefiniteness puts these vectors in the radical of $Q_X$. With the shear invariance of the intrinsic denominator (Proposition~\ref{prop:rp-intrinsic-denominator}), the quadratic pole form annihilates the full kernel of $\Jac_X$ and has rank at most eight. \subsection{The auxiliary quadratic family} Let $D(X,Y)$ be the intrinsic reduced denominator of Proposition~\ref{prop:rp-intrinsic-denominator}. Its fibre at a generic anchor of the open set in Proposition~\ref{prop:dq-exclusion} is $Q_X^2$, so by Proposition~\ref{prop:rp-intrinsic-denominator} it has bidegree $(4,4)$ and no factor in one slot only, so each irreducible factor of $D$ in $\R[X,Y]$ is primitive in $Y$, and Gauss's lemma identifies the factorization with the one over $\R(X)[Y]$. Distinct irreducible factors are coprime over $\R(X)$ and squarefree over its algebraic closure, the characteristic being zero; after a generic linear change of the $Y$-coordinates their resultants and discriminants in one coordinate are nonzero polynomials, which stay nonzero off a proper algebraic set of anchors. On the nonempty Zariski-open set of anchors where these resultants and discriminants are nonzero and the fibre of $D$ is the reduced denominator (Proposition~\ref{prop:rp-intrinsic-denominator}), a set that meets that open set, each factor fibre divides $Q_X^2$ and is squarefree, hence is proportional to the irreducible quadratic $Q_X$, and coprimeness leaves one factor, with exponent two. Hence \begin{equation}\label{eq:dq-global-square} D=c\,\mathcal Q(X,Y)^2 \end{equation} with $\mathcal Q$ real of bidegree $(2,2)$ and $c\ne0$. The two shear identities of $D$ pass to $\mathcal Q$: their squares agree in the polynomial ring with the shear parameter, the domain property gives a common sign, and parameter zero makes it positive. The shears generate $(X,Y)\mapsto(Y,-X)$, and the even bidegrees give $\mathcal Q(Y,X)=\mathcal Q(X,Y)$, $\mathcal Q(X,Y+tX)=\mathcal Q(X,Y)$ and $\mathcal Q(X,X)=0$. Write $\mathcal Q(X,Y)=\langle A_XY,Y\rangle$; then $A_X$ is a symmetric quadratic family with $A_XX=0$. Anchors at which Proposition~\ref{prop:rp-denominator-types} applies and the fibre of $D$ is $D_X$ are dense. At each of them $c\,\mathcal Q(X,\cdot)^2=D_X$, and among the factor types of Proposition~\ref{prop:rp-denominator-types} the only square is the square of an irreducible quadratic, so $\mathcal Q(X,\cdot)$ is an irreducible positive semidefinite quadratic with a double pole and the argument after \eqref{eq:dq-determinant} applies at $X$. Semidefiniteness persists by continuity. Fix one such anchor $X_0$ and the sign of $\mathcal Q$ so that $\mathcal Q(X_0,\cdot)\succeq0$; symmetry gives $\mathcal Q(Y,X_0)=\mathcal Q(X_0,Y)>0$ for $Y$ outside its kernel, so $A_Y\succeq0$ on a dense set and then everywhere. The argument after \eqref{eq:dq-determinant} and continuity of the kernel projector give \begin{equation}\label{eq:dq-aligned-family} \ker\Jac_X\subset\ker A_X,\qquad A_X\succeq0. \end{equation} \begin{lemma}\label{lem:dq-common-coefficient} Let $A_z$ be a symmetric quadratic family with $\operatorname{Im}A_z\subset\operatorname{Im}\Jac_z$ for all real $z\ne0$. For the anchor $X$ and, as a rational identity, for $Y\in X^\perp$, \[ A_{sX+Y}=(sM_X+M_Y)C(sM_X+M_Y)^T,\qquad C=M_X^TA_XM_X, \] with $C$ constant, $C\succeq0$ if $A\succeq0$, and \begin{equation}\label{eq:dq-commutator} \Omega HC-CH\Omega\quad\text{polynomial}. \end{equation} \end{lemma} \begin{proof} For a generic partner $Y$ the matrix $\widehat M=[M_X,M_Y]$ is injective, and $A_{sX+tY}=\widehat M\mathcal M(s,t)\widehat M^T$ for a unique symmetric quadratic block matrix $\mathcal M$. Its values at the axes give \[ \mathcal M=\begin{bmatrix}s^2C+st\Theta_{11}&st\Theta_{12}\\st\Theta_{12}^T&t^2C_1+st\Theta_{22}\end{bmatrix}. \] Since $\operatorname{Im}A_{sX+tY}\subset\operatorname{Im}\widehat M[sI,tI]^T$, we get $[-tI,sI]\mathcal M=0$ for generic real $s,t$, hence identically, and comparing coefficients gives $\Theta_{11}=\Theta_{22}=0$ and $\Theta_{12}=C_1=C$. This holds on the generic pair locus and then rationally. The mixed coefficient of $A_{sX+Y}$ makes $FC+CF^T$ polynomial, and $F=F_0+\Omega H$ gives \eqref{eq:dq-commutator}. \end{proof} By \eqref{eq:dq-aligned-family} the lemma applies here, with $\rank C=\rank Q_X$, since the fibre defining $A_X$ is proportional to $Q_X$. \subsection{The polar part and total isotropy} \begin{lemma}\label{lem:dq-block-normal-form} Let $C\succeq0$ have rank $r$, put $r'=8-r$, and let $\Omega$ be skew rational with $P\Omega$ and $\Omega HC-CH\Omega$ polynomial. After subtracting a polynomial skew matrix from $\Omega$, \begin{equation}\label{eq:dq-exact-kernel} \Omega HC=CH\Omega,\qquad P\Omega\ \text{polynomial}. \end{equation} After a constant change $S$ of coefficient coordinates, $C'=S^{-1}CS^{-T}$, $H'=S^THS$, $\Omega'=S^{-1}\Omega S^{-T}$, $P'=PS$ (this keeps skewness, \eqref{eq:dq-exact-kernel}, generic injectivity of column subpencils, and the lower output frame), on spaces $U_1,U_2,U_3,U_4$ of dimensions $d,n_0,n_0,\ell$ with $d=r-n_0$, $\ell=r'-n_0$, $n_0\le\min(r,r')$, \begin{equation}\label{eq:dq-form-normalization} C=\operatorname{diag}(I_d,I_{n_0},0_{n_0},0_\ell),\qquad H=\begin{bmatrix} H_0&0&0&0\\0&0&I_{n_0}&0\\0&I_{n_0}&0&0\\0&0&0&H_1 \end{bmatrix}, \end{equation} \begin{equation}\label{eq:dq-whole-blocks} \Omega=\begin{bmatrix} Z&0&0&0\\ 0&B_0&\Psi&T_0\\ 0&\Psi&0&0\\ 0&-T_0^T&0&V_0 \end{bmatrix},\qquad ZH_0=H_0Z, \end{equation} with $H_0,H_1$ nondegenerate, $Z,B_0,\Psi,V_0$ skew, and empty blocks allowed. \end{lemma} \begin{proof} The constant linear map $\mathcal L(\Omega_1)=\Omega_1HC-CH\Omega_1$ on skew matrices is injective on a constant complement of its kernel, so the component of $\Omega$ in that complement is polynomial; subtracting it gives \eqref{eq:dq-exact-kernel}. Normalize $C=\operatorname{diag}(I_r,0_{r'})$ and let $n_0$ be the nullity of $H$ on the first $r$-space; split that space orthogonally into its nondegenerate part and radical. The radical pairs with the last $r'$-space with row rank $n_0$, since $H$ is nondegenerate, so $n_0\le r'$. Changes $\left[\begin{smallmatrix}O&T\\0&N\end{smallmatrix}\right]$ with $O$ orthogonal and $N$ invertible preserve $C$: the first $d$ rows of $T$ remove the cross terms of $H_0$, $N$ normalizes the radical pairing to $[I_{n_0},0]$, and the remaining rows of $T$ remove the terms on the paired $n_0$-space. This gives \eqref{eq:dq-form-normalization}, and nondegeneracy of $H$ gives that of $H_1$; comparing blocks in $\Omega HC=CH\Omega$ gives \eqref{eq:dq-whole-blocks}; the two copies of $\Psi$ have the same sign because $\Psi^T=-\Psi$. \end{proof} \begin{proof}[Completion of Proposition~\ref{prop:dq-exclusion}] Apply Lemma~\ref{lem:dq-block-normal-form} with $r=\rank Q_X$; the subtracted matrix is polynomial, so $\Omega$ keeps its double pole. Diagonalize $H_0$ orthogonally. Since $ZH_0=H_0Z$ by \eqref{eq:dq-whole-blocks}, $Z$ splits into skew blocks on the nonzero eigenspaces of $H_0$, each definite for the indefinite form $H$ and so of dimension at most seven; the skew block on the last three spaces of \eqref{eq:dq-whole-blocks} has dimension $r'+n_0=8-d$, also at most seven if $d>0$. These are exact constant block decompositions, so the corresponding columns of $P\Omega$ are polynomial products of linear, generically injective column subpencils with the blocks, and Lemma~\ref{lem:dq-block-degree} excludes a double pole in every block of dimension at most seven. If $d>0$, every block has dimension at most seven, so none carries the double pole of $\Omega$; hence $d=0$, that is \begin{equation}\label{eq:dq-isotropic-range} CHC=0, \end{equation} so the range of $C$ is totally $H$-isotropic and $r$ is at most the Witt index of $H$; hence $r\le4$. \emph{Ranks two and three.} Since $Q_X$ is real irreducible and semidefinite, $r\ge2$. With $d=0$ we have $n_0=r$, $\ell=8-2r$ and \[ \Omega=\begin{bmatrix}B_0&\Psi&T_0\\ \Psi&0&0\\ -T_0^T&0&V_0\end{bmatrix}. \] Write $P=[P_2,P_3,P_4]$ by the column blocks on $U_2,U_3,U_4$; the middle columns of $P\Omega$ are $P_2\Psi$. For $r=2,3$ a pole of the skew block $\Psi$, even a simple one, would need $r\ge4$ by Lemma~\ref{lem:dq-block-degree}, and all its denominators divide $Q_X^2$, so $\Psi$ is polynomial. Subtracting it leaves the double pole on the first and last spaces, of total dimension $8-r<8$, which the same lemma excludes. \emph{Rank four.} Now $H$ has signature $(4,4)$ and the normalization gives \[ H'=\Lambda'=\begin{bmatrix}0&I_4\\I_4&0\end{bmatrix},\quad C'=\begin{bmatrix}I_4&0\\0&0\end{bmatrix},\quad \Omega'=\begin{bmatrix}B_0&\Psi\\\Psi&0\end{bmatrix},\quad P'=[P_2,P_3]. \] Polynomiality of the four columns $P_2\Psi$ excludes a double pole of $\Psi$; if $\Psi$ were polynomial, subtracting it would leave a double pole in a four-dimensional block, again excluded. Thus $\Psi$ has a simple $Q_X$-pole, so every maximal four-minor of $P_2$ is divisible by $Q_X^2$. These minors have degree four, so they are constant multiples of $Q_X^2$, not all zero; their Pl\"ucker vector gives one fixed real four-space $T$ with $\operatorname{Im}P_2(Y)\subset T$, first generically and then for all $Y$ by polynomial continuation. Choose $0\ne\eta\in T^\perp$ in the original Euclidean lower output coordinates and put $z=K_X\eta$. Only coefficient coordinates were changed, so $\eta^TP'(Y)=(0,b(Y)^T)$ and $P'\Lambda'P'^T=P\Lambda P^T$. The hyperbolic $\Lambda'$ gives $\langle\Jac_{tX+Y}z,z\rangle=\eta^TP'(Y)\Lambda'P'(Y)^T\eta=0$ for all real $t,Y$; these are polynomial identities, so no direction is lost at the frame singularities. The symmetry $\langle\Jac_zV,V\rangle=\langle\Jac_Vz,z\rangle$ of the curvature tensor gives $\langle\Jac_zV,V\rangle=0$ for every real $V$, and polarization gives $\Jac_z=0$, contradicting rank eight at $z\ne0$. So each of the ranks $r=2,3,4$ is excluded, which proves Proposition~\ref{prop:dq-exclusion}. \end{proof}