\providecommand{\C}{\mathbb C} \section{Assembly of the proof of Theorem~A}\label{sec:assembly} \subsection{The anchor set and the pole alternative} Fix $R$ as in Theorem~\ref{thm:conjA}. If $R$ has two reduced eigenvalues, Theorem~\ref{thm:A} applies. Theorems~\ref{thm:A}, \ref{thm:D} and~\ref{thm:E} classify twenty of the thirty-six multiplicity patterns on $\R^{16}$ that contain a multiplicity $7$ or $8$ (the two-eigenvalue pattern $(8,7)$ and nineteen with three or more eigenvalues), leaving exactly sixteen: \[ \begin{gathered} (8,4,3),\ (8,3,3,1),\ (8,3,2,2),\ (8,3,2,1,1),\ (8,3,1^4);\\ (7,5,3),\ (7,4,4),\ (7,4,3,1),\ (7,4,2,2),\ (7,4,2,1,1),\ (7,4,1^4),\\ (7,3,3,2),\ (7,3,3,1,1),\ (7,3,2,2,1),\ (7,3,2,1^3),\ (7,3,1^5). \end{gathered} \] Shift the eigenvalue of multiplicity $8$ or $7$ to zero. The shifted family has constant rank $\nu=7$ or $\nu=8$ on $\R^{16}\setminus0$, and its nonzero spectral matrix $\Lambda$ is nonscalar. If $\Lambda$ is definite, Theorem~\ref{thm:extreme} shows that $R$ is Clifford; from now on $\Lambda$ is indefinite. The construction of Lemma~\ref{lem:rationalpartner} applies at an anchor $x$ whenever the generic-pair conditions of \cite[Section~4, Lemmas~2(1) and~3, pp.~7--9 of arXiv:math/0204258v2]{Nik2005} hold for $(x,y)$ with $y$ in a dense subset of $x^\perp$; the condition $\operatorname{Im}\Jac_x\cap\operatorname{Im}\Jac_y=0$ is the nonvanishing of the sum of squares of the $2\nu$-minors of the $16\times32$ matrix $[\Jac_x\ \Jac_y]$, that is, $\rank[\Jac_x\ \Jac_y]=2\nu$, a polynomial condition $g(x,y)\ne0$. Substituting $y=(\norm{x}^2I-xx^T)z$ and expanding in $z$, the anchors at which $g(x,\cdot)$ vanishes identically form a real algebraic set, proper because a good generic pair exists. Remove it, together with the proper algebraic exceptional loci used in the preceding sections: the non-generic anchors of Lemma~\ref{lem:rp-generic-triples} and Propositions~\ref{prop:rp-denominator-types} and~\ref{prop:rp-intrinsic-denominator} (a semialgebraic set of dimension less than sixteen, hence inside a proper algebraic set) and the loci used for spreading out geometric factors (Remark~\ref{rem:asm-galois}). The result is a nonempty Euclidean-open cone $U_0\subset\R^{16}\setminus0$; its points are the \emph{regular anchors} of Section~\ref{sec:coherent-transpose} (the construction at $x$ is understood after normalizing to $x/\norm x$). For $x\in U_0$ we write $F_x$ for the rational upper block and $P_x$ for the linear lower block, of size $(16-\nu)\times\nu$. By Lemma~\ref{lem:cp-pole-free}, if $F_x$ has no pole for one $x\in U_0$, then $R$ is Clifford. So we may assume that $F_x$ has a pole for every $x\in U_0$; the rest of the proof shows that this again leads to a Clifford tensor. \subsection{A finite semialgebraic case partition} At each $x\in U_0$ the pole data of $F_x$ consist of finitely many discrete invariants: the real factorization type of the reduced denominator and its pole orders; the ranks of the skew residues; the real residue support $d$; the residue-variety dimension $\alpha$ and the generic transpose rank $s$; the number, degrees and dimensions of the complex spans of the residue images; the relative multiplicities of the coefficient pencil and the signatures involved; and whether the transpose Gram vanishes on the singular stratum. They do not depend on the choice of orthonormal eigenframes at $x$: two choices differ by constant orthogonal block transformations commuting with $\Lambda$, which preserve every rank, dimension, support, degree and image inclusion. \begin{lemma}[Real-open branch]\label{lem:asm-partition} There is a finite partition $U_0=\mathcal U_1\sqcup\cdots\sqcup\mathcal U_N$ into semialgebraic sets on each of which every invariant above is constant. At least one $\mathcal U_j$ has nonempty Euclidean interior. \end{lemma} \begin{proof} All objects have bounded degree in $(x,y)$, so each invariant above takes only finitely many values. The geometric invariants (number, degrees, dimensions and spans of the residue images, $\alpha$ and $s$) are constructible functions of $x$ \cite[\S9.7]{EGAIV3}, as in Remark~\ref{rem:asm-galois}, and a constructible set defined over $\R$ has a semialgebraic set of real points. Include the image and kernel frames at $x$ as auxiliary real variables, constrained by the polynomial orthonormality and spectral equations; by frame independence, existential and universal quantification over the frames give the same predicate. Ranks and signatures are sign conditions on minors; a factorization of bounded degree is a system of coefficient equations with auxiliary real and imaginary coefficients; the absence of real points, the positive pencil member and its relative multiplicities are quantified polynomial conditions. So each value of each invariant is a first-order formula over $\R$ in $x$, and by Tarski--Seidenberg elimination \cite[Chapter~2]{BCR1998} each such set is semialgebraic and the common refinement is finite. Outside the union of the zero sets of the finitely many nonzero polynomials in quantifier-free descriptions of the $\mathcal U_j$, membership in each $\mathcal U_j$ is locally constant; a finite union of proper real zero sets contains no nonempty open set, so some point of $U_0$ has a neighbourhood inside one $\mathcal U_j$. \end{proof} \noindent The branches need not be Zariski open, and signature boundaries may form lower-dimensional strata; the proof never uses them. \begin{remark}[Galois arguments]\label{rem:asm-galois} Several branch results use the action of a Galois group on the geometric factors of the pole data over $\R(x)$. Irreducibility over $\R(x)$, or a particular orbit, need not survive at a real anchor ($y^2-x_1$ is irreducible over $\R(x)$ but splits over $\R$ when $x_1>0$). What persists, after a finite algebraic cover and the removal of a proper algebraic subset, are the geometric components and their algebraic invariants (ranks, degrees, spans and $(\alpha,s)$): the number of geometric irreducible components of the fibres of a morphism of finite type, and these invariants of the components, are constructible functions on the base \cite[\S9.7]{EGAIV3}. So if Galois transitivity forces such an invariant to be equal on all conjugate geometric factors, the equality persists at the specialized anchors, even where real conjugation orbits split; real splittings and signatures are recorded by Lemma~\ref{lem:asm-partition}. This is the only form in which Galois transitivity is used. \end{remark} \subsection{The cases of the partition} Every branch is covered by one of the results below. Each either classifies $R$ directly, contradicts the branch, or gives \emph{full transpose alignment} throughout the real-open region of the branch: the full transpose pencil has image $E_x=\operatorname{Im}\Jac_x$ for every real coefficient $a\ne0$, at every anchor $x$ of the region. \begin{enumerate}[label=(L\arabic*),leftmargin=*] \item $\nu=7$: there is no pole at any anchor (Proposition~\ref{prop:nu7-no-poles}). \item $\nu=8$, pole orders. The reduced denominator $D$ satisfies $D^2\mid\det P_x$ and has no real linear or odd-degree factor, so it is a quadratic $Q$, its square, a product $Q_1Q_2$ of distinct quadratics, or a real-irreducible quartic $f$ (Proposition~\ref{prop:rp-denominator-types}). Double quadratic poles are excluded (Proposition~\ref{prop:dq-exclusion}), as is a simple quadratic pole with residue rank four (Proposition~\ref{prop:sr4-exclusion}). A quartic residue has rank two by Lemma~\ref{lem:cp-dvr-budget}, and for $Q_1Q_2$ the same lemma, $Q_1^{r_1}Q_2^{r_2}\mid\det P_x$, gives $r_1=r_2=2$. So every remaining pole is simple, with residue rank two. \item Proper real support $d\le7$. For a quadratic component, $Q^2$ divides the maximal $d$-minors, so $d\ge4$; $d=4,5$ are excluded componentwise (Lemma~\ref{lem:si-small-support}). For a sole quadratic, $d=6$ is excluded by Proposition~\ref{prop:ps-six-exclusion}, and both $d=7$ alternatives, $s=4$ and $s=5$, by Proposition~\ref{prop:ps-seven-sole}. For two quadratics, a component with $d=6$ is excluded by Proposition~\ref{prop:ps-six-exclusion}, and $d=7$ by Proposition~\ref{prop:ps-two-seven}. A real-irreducible quartic pole always has $d=8$ (Lemma~\ref{lem:si-quartic-support}). \item Full support. Incidence gives $4\le s\le\alpha\le5$; the pair $(\alpha,s)=(5,4)$ gives full transpose alignment (Proposition~\ref{prop:si-five-four}). In the remaining cases $\alpha=s$, and Lemma~\ref{lem:si-saturated-fibres} makes distinct pole components have distinct residue images. \item Two quadratic poles. Proposition~\ref{prop:aq-two-structure} leaves two geometric quadratics or four geometric lines in one Galois orbit, so all geometric components have one type $(\alpha,s)$ (Remark~\ref{rem:asm-galois}). Type $(5,5)$ is excluded by the divisor budget (Proposition~\ref{prop:db-twoQ}), and type $(4,4)$ is aligned by Proposition~\ref{prop:u4-degree-budget}, whose hypotheses hold by Remark~\ref{rem:r4-generic-finish}. \item A single real pole of type $(4,4)$. By Remark~\ref{rem:r4-generic-finish}, $G$ vanishes on every image of type $(4,4)$, and full real support gives full complex span. A real geometrically integral image of full span has degree $4$, $6$ or $8$, and each is treated in Proposition~\ref{prop:r4-real-image}. A conjugate pair of images aligns (Proposition~\ref{prop:u4-conjugate-pair}). \item A single real pole of type $(5,5)$ with one real integral image $V(h,q)$, where $q$ is positive and the maximal relative multiplicity of the pencil $(h,q)$ is at most five (Propositions~\ref{prop:db-single} and~\ref{prop:pf-positive-model}). The maximal multiplicity is at most three (Propositions~\ref{prop:pf-factorial-cone} and~\ref{prop:pf-factorial-normal-form}), five (Propositions~\ref{prop:zs-vanishing} and~\ref{prop:sie-extraction}, using Lemma~\ref{lem:sie-five}), or four (Propositions~\ref{prop:zs-vanishing} and~\ref{prop:sie-extraction}, using Lemmas~\ref{lem:sie-four} and~\ref{lem:sie-four-four}). The normal forms produced there align: unbalanced forms by Proposition~\ref{prop:zs-unbalanced}, balanced forms by Proposition~\ref{prop:balanced-spectral-applications}, which covers every relative multiplicity at most five; the extraction uses Lemma~\ref{lem:cl-c}, Proposition~\ref{prop:zs-vanishing} and Lemma~\ref{lem:sie-node}. \item A single real pole of type $(5,5)$ with a conjugate pair of images $V(h,\ell)$, $V(h,\overline\ell)$ (Proposition~\ref{prop:db-single}, Lemma~\ref{lem:db-definite} and Proposition~\ref{prop:db-positive-model}). A transpose Gram vanishing on $\Sigma=V(h,\ell)\cap V(h,\overline\ell)$ gives alignment (Proposition~\ref{prop:m6-zero-intersection}), and so does a constant column (Proposition~\ref{prop:m6-constant-column}, using Lemma~\ref{lem:m03-ten-space}). Otherwise Proposition~\ref{prop:m6-linear-extraction} supplies complex linear columns, and Proposition~\ref{prop:cg-real-gluing} gives a global real balanced form, which aligns by Proposition~\ref{prop:balanced-spectral-applications}. \end{enumerate} The list is exhaustive. Items (L2)--(L4) reduce every pole at $\nu=8$ to simple poles with rank-two residues, full real support, and type $(\alpha,s)=(4,4)$ or $(5,5)$. The denominator has one real pole component ($Q$ or $f$) or two ($Q_1Q_2$), the latter being (L5). A single real pole component is geometrically irreducible or splits over $\C$ into one conjugate pair, so it has one geometrically integral real image or a conjugate pair of images; the two conjugate components of a split real pole have the same $(\alpha,s)$, since $B$ is real. With the two types this gives four alternatives: the two of (L6), and (L7) and (L8); in (L7) the maximal relative multiplicity takes one of the three listed values. \subsection{Proof of Theorem~\ref{thm:conjA}} \begin{proof} By the reductions above we may assume that $\Lambda$ is indefinite. Suppose that $R$ is not Clifford. Then $F_x$ has a pole at every $x\in U_0$, and $\nu=8$ by (L1); the cases that conclude ``Clifford'' now give contradictions. By Lemma~\ref{lem:asm-partition} some branch $\mathcal U_j$ contains a nonempty open set $\mathcal U'$ of anchors. If the result of (L1)--(L8) covering $\mathcal U_j$ shows that $R$ is Clifford, or that no anchor has the invariants of $\mathcal U_j$, we have a contradiction. Otherwise it gives full transpose alignment: at every $x\in\mathcal U'$ the full transpose pencil has image $E_x$ for every real $a\ne0$. By Lemma~\ref{rem:m03-scope}, this gives $\Jac_y z=0$ for all $y,z\in\ker\Jac_x$, $x\in\mathcal U'$. Lemma~\ref{lem:m03-closed-kernel} applies on the open set $\mathcal U'$ and makes the full Gram matrix of the coherent partner radial, hence polynomial. Since $\Lambda$ is nonscalar, Proposition~\ref{prop:gramconditional}(2) shows that $R$ is Clifford, a contradiction. \end{proof}