\section{The divisor budget for five-dimensional residue images} \label{sec:db} Throughout this section $\nu=8$, the anchor is regular, and $B(a)$ is the actual real linear transpose pencil, with $8$ coefficient variables $a=(a_1,\ldots,a_8)$. By Lemma~\ref{lem:cp-L20}, $B(a)$ has rank $8$ for every real $a\ne0$. Let $h(a)=a^THa$ be the real coefficient form; it is nondegenerate and indefinite. Every residue image $A$ lies in the projective quadric $V(h)\subset\mathbb P^7_{\mathbb C}$, by the isotropy in \eqref{eq:cp-residue}. We consider poles of type $(\alpha,s)=(5,5)$: each geometric pole component has an irreducible closed residue image $A$ of dimension five, and $B$ has generic rank five on $A$. Distinct geometric pole components have distinct images (Lemma~\ref{lem:si-saturated-fibres}). Since rank is lower semicontinuous, $\rank B(a)\le5$ at \emph{every} point of $A$. Hence $A$ has no real point. The real residue support is the smallest real subspace $U\subset\R^8$ whose complexification contains every residue image. If a real-defined image $A$ lies in a complex hyperplane $\{c\cdot a=0\}$, it also lies in $\{\overline c\cdot a=0\}$, hence in a real hyperplane. So a real-defined image of a component with real support $8$ spans $\mathbb P^7$. \subsection{The budget} Let $S=\mathbb C[a_1,\ldots,a_8]/(h)$ with its standard grading. By the Klein--Nagata theorem \cite[Theorem~8.2, p.~26]{Samuel1964}, a nondegenerate quadratic form in at least five variables over a field of characteristic not two gives a factorial quotient; thus $S$ is a UFD. The affine cone over a five-dimensional irreducible $A\subset V(h)$ is a height-one prime of $S$, so its ideal is generated by a homogeneous irreducible $f_A$ of some degree $e_A\ge1$. If $A$ spans $\mathbb P^7$, then $e_A\ge2$. \begin{lemma}[Divisor budget]\label{lem:db-budget} Let $A_1,\ldots,A_k$ be distinct residue images of type $(5,5)$. Then \[ e_{A_1}+\cdots+e_{A_k}\le2 . \] \end{lemma} \begin{proof} Over the fraction field of $S$ the matrix $B$ has rank $8$: the real cone of $h$ contains nonzero points, where $B$ has rank $8$, so some maximal minor of $B$ is nonzero in $S$. Fix $i$ and localize $S$ at the height-one prime $(f_{A_i})$; the result is a discrete valuation ring. The reduction of $B$ has rank five at the generic point of $A_i$, so in the local Smith form of the $15\times8$ matrix $B$ (its image lies in $X^\perp$) at least three of the eight invariant factors lie in the maximal ideal. Hence every maximal minor of $B$ is divisible in $S$ by $f_{A_i}^3$. The $f_{A_i}$ are pairwise nonassociate primes, so every maximal minor is divisible by $\prod_if_{A_i}^3$, of degree $3\sum_ie_{A_i}$. A maximal minor has degree $8$ and some maximal minor is nonzero in $S$. Therefore $3\sum_ie_{A_i}\le8$. \end{proof} \subsection{Consequences} \begin{proposition}[Two quadratic poles]\label{prop:db-twoQ} Suppose that the reduced denominator is $Q_1Q_2$ with two distinct real quadratic factors, that every real residue support is $8$, and that all geometric pole components have type $(5,5)$. This configuration does not occur. \end{proposition} \begin{proof} Each real quadratic factor contributes at least two to the sum of the image-generator degrees. If it is geometrically irreducible, its single image is real-defined and has full span, so $e\ge2$. If it splits, it has two distinct conjugate geometric components with distinct images, each with $e\ge1$; no individual full span is needed. Images belonging to different pole components are also distinct by Lemma~\ref{lem:si-saturated-fibres}. Thus the two real quadratic factors contribute a total of at least four, contrary to Lemma~\ref{lem:db-budget}. This count also covers the mixed pattern of one geometrically irreducible and one split factor. \end{proof} \begin{proposition}[A single real pole]\label{prop:db-single} Suppose that the pole has one real component, with full real support and type $(5,5)$. \begin{enumerate} \item If its image $A$ is real-defined and geometrically integral, then $A=V(h,f)$ for a real quadric $f$; it is a complete intersection of degree four with no real point. \item Otherwise there are exactly two images, $A$ and $\overline A$, and $A=V(h,\ell)$, $\overline A=V(h,\overline\ell)$ for a complex linear form $\ell$ that is not proportional to a real form. \end{enumerate} \end{proposition} \begin{proof} A real irreducible polynomial is geometrically irreducible or splits into one pair of complex conjugate factors. (1) Full real support gives full complex span (see the start of this section), so $e_A\ge2$, and Lemma~\ref{lem:db-budget} gives $e_A=2$. Thus $A=V(h,f_A)$ with $f_A$ of degree two. Since $A$ is real-defined, $\overline{f_A}$ generates the same prime, so $\overline{f_A}=cf_A$ for a nonzero constant $c$, the only units of $S$ in degree zero. Applying conjugation twice gives $|c|=1$; write $c=\omega/\overline\omega$ and replace $f_A$ by $\omega f_A$, which is fixed by conjugation. Any real lift $f$ of this class defines the same complete intersection. Its degree is $2\cdot2=4$, and it has no real point by the remark before the lemma. (2) The two images are distinct by Lemma~\ref{lem:si-saturated-fibres} and conjugate, so $e_A=e_{\overline A}$, and Lemma~\ref{lem:db-budget} gives $e_A=e_{\overline A}=1$. Thus $A=V(h,\ell)$ with $\ell$ linear. If $\ell$ were proportional to a real form, $A$ would be real-defined and equal to $\overline A$. \end{proof} \subsection{The conjugate pair: definiteness and the positive model} We now normalize case~(2) of Proposition~\ref{prop:db-single}. Write $\ell=\ell_1+i\ell_2$ with real linear forms $\ell_1,\ell_2$; they are linearly independent. Put \[ W=\{a\in\R^8:\ell_1(a)=\ell_2(a)=0\},\qquad r=\ell_1^2+\ell_2^2=\ell\,\overline\ell . \] Then $\dim W=6$, and $r$ is a real positive semidefinite form of rank two. \begin{lemma}\label{lem:db-definite} The restriction $h|_W$ is definite. Consequently the signature of $h$ is $(6,2)$ or $(7,1)$, up to an overall sign. \end{lemma} \begin{proof} A real point of $V(h,\ell)$ is a real $a\ne0$ with $h(a)=0$ and $\ell(a)=0$, that is, $\ell_1(a)=\ell_2(a)=0$. So the real points of $A=V(h,\ell)$ are exactly the real zeros of $h$ on $W$. Since $A$ has no real point, $h$ has no nonzero real zero on $W$, and a real quadratic form without nontrivial real zeros is definite. A nondegenerate form that is definite on a six-dimensional subspace has at least six positive or at least six negative squares; since $h$ is indefinite, its signature is $(6,2)$ or $(7,1)$ up to sign. \end{proof} \begin{proposition}[Positive model of the conjugate pair]\label{prop:db-positive-model} There are a positive definite real quadratic form $q$ and a real constant $\lambda\ne0$ such that $\rho=h-\lambda q$ is a semidefinite form of rank two with radical $W$, whose restriction to $W^{\perp_q}$ is definite. In $q$-orthonormal coordinates $a=(x,z)$ with $x\in W$ and $z\in W^{\perp_q}$, \[ q=|x|^2+|z|^2,\qquad h=\lambda q+\rho(z), \] and \[ A\cup\overline A=V(h,r)=V(h,q)=V(q,\rho). \] In particular the pencil spanned by $h$ and $q$ contains the rank-two member $\rho$, and $\lambda$ is a relative eigenvalue of $h$ with respect to $q$ of multiplicity six. \end{proposition} \begin{proof} The forms $h$ and $r$ have no common nonzero real zero: such a zero would lie in $W$ and be a real zero of $h$ there, which is excluded by Lemma~\ref{lem:db-definite}. The joint range of $(h,r)$ on the real unit sphere of $\R^8$ is compact, and it is convex by Brickman's theorem \cite[Theorem~1.1]{MartinezLegaz2005}. It avoids the origin. If $v$ is its point closest to the origin, then $\langle v,w\rangle\ge\langle v,v\rangle>0$ for every point $w$ of the range. This gives a positive definite combination $q=\alpha h+\beta r$. Here $\alpha\ne0$, because $r$ has rank two, and $\beta\ne0$, because $h$ is indefinite. Put $\lambda=1/\alpha$ and $\rho=-(\beta/\alpha)r=h-\lambda q$. Then $\rho$ is a nonzero multiple of $r$, so it is semidefinite of rank two with radical $W$, and definite on $W^{\perp_q}$. Choose $q$-orthonormal coordinates $x$ on $W$ and $z$ on its $q$-orthogonal complement. The forms $\ell_1,\ell_2$ vanish on $W$, so they, and hence $\rho$, depend on $z$ alone. This gives the displayed model. Finally $V(h,q)=V(h,\alpha h+\beta r)=V(h,r)$ since $\beta\ne0$, and $V(h,r)=V(h,\ell\overline\ell)=V(h,\ell)\cup V(h,\overline\ell)$. Also $V(h,q)=V(q,h-\lambda q)=V(q,\rho)$. On $W$ the relative eigenvalue of $h$ with respect to $q$ is $\lambda$, and on $W^{\perp_q}$ the form $h-\lambda q=\rho$ is nondegenerate, so the multiplicity of $\lambda$ is exactly six. \end{proof} \begin{remark}\label{rem:db-scope} Proposition~\ref{prop:db-positive-model} is the input \eqref{eq:m6-positive-model} of Section~\ref{sec:m6-preliminary}. The union $A\cup\overline A$ is a \emph{reducible} complete intersection, so the factorial and normality arguments for a geometrically integral image (Section~\ref{sec:positive-factorial}) do not apply to it; the conjugate pair is treated in Section~\ref{sec:m6-preliminary} and Section~\ref{sec:conjugate-gluing}. Each component $V(h,\ell)$ is a quadric of rank six or seven in its hyperplane, since restricting the rank-eight form $h$ to a hyperplane lowers its rank by at most two; in particular each component is irreducible and spans its hyperplane. \end{remark}