\section{An extreme large Jacobi root} \begin{mainthm}\label{thm:extreme} Let $R$ be an Osserman algebraic curvature tensor on $\R^{16}$ with at least three distinct reduced Jacobi eigenvalues. If an eigenvalue of multiplicity seven or eight is the smallest or the largest reduced eigenvalue, then $R$ is Clifford. \end{mainthm} The two-root case has the Clifford and Cayley alternatives of \classAtheorem. The qualification on the number of roots is necessary: the Cayley tensor has two extreme roots. We first establish the tools used in the proof. Let $\Pi(x)=\norm x^2I-xx^T$. Throughout this section a constant-curvature shift has been made, so that the reduced Jacobi operator has a zero root and nonzero rank $\nu\le8$. Let $\Lambda$ be diagonal with the nonzero Jacobi roots for unit vectors, including their multiplicities. \begin{proposition}\label{prop:gramconditional} Let $R$ be an Osserman algebraic curvature tensor on $\R^{16}$ as above. \begin{enumerate} \item If $\Jac_x=M_x\Lambda M_x^T$ has a global linear factor $M_x:\R^\nu\to\R^{16}$, injective for all $x\ne0$, then $R$ is Clifford. \item Suppose instead that $\Jac_z=M_z\Lambda M_z^T$ has a rational homogeneous factor of degree one with the following coherence property. There is a real unit vector $X$ and an orthonormal eigenframe $M_X$ such that, for $Y\perp X$, \begin{equation}\label{eq:shear} M_{sX+Y}=sM_X+M_Y \end{equation} as a rational identity. If the \emph{full} Gram matrix $\Phi(z)=M_z^TM_z$ on $\R^{16}$ is polynomial, then $R$ is Jacobi-proportional. Consequently it is Clifford or belongs to the Cayley affine family. If $\Lambda$ is nonscalar, it is Clifford. \end{enumerate} \end{proposition} The last classification uses \classcor. Polynomiality only on $X^\perp$ is not the hypothesis in part (2): in \eqref{eq:shear} the full Gram is \[ \Phi(sX+Y)=s^2I+s(M_X^TM_Y+M_Y^TM_X)+M_Y^TM_Y, \] so its mixed coefficient must also be polynomial. Neither part asserts that the required global factor or polynomial Gram exists in the indefinite case. \subsection{Quadratic Gram rigidity} \begin{lemma}\label{lem:gramrigidity} Let $\Lambda$ be invertible symmetric of size $\nu\le8$. Suppose $\Phi(x)$ is symmetric homogeneous quadratic on $\R^{16}$, positive definite for $x\ne0$, and $\Lambda\Phi(x)$ has the spectrum of $\norm x^2\Lambda$, with multiplicities. If $\Phi(x_0)=I$ at a unit vector $x_0$, then $\Phi(x)=\norm x^2I$ for every $x$. \end{lemma} \begin{proof} The argument refines \cite[Section~5, Lemmas~5--8, pp.~11--15 of the preprint]{Nik2005}, whose hypothesis $n>(\nu+1)^2/4$ fails for $n=16$, $\nu=7,8$; the remaining pairs are closed by the topological count below. We may diagonalize $\Lambda$ orthogonally. Choose a root $\lambda$ of multiplicity $m$ whose reciprocal is extremal, and set $q=\norm x^2$, \[ Q_\lambda(x)=\lambda q\Lambda^{-1}-\Phi(x). \] For unit $x$, put $H=\Phi(x)^{1/2}$. Since $H\Lambda H$ is symmetric with the spectrum of $\Lambda$, write $H\Lambda H=O\Lambda O^T$. Then $C=O^TH$ satisfies $C\Lambda C^T=\Lambda$, $\Phi=C^TC$, and \[ Q_\lambda(x)=C^T(\lambda\Lambda^{-1}-I)C. \] Thus $Q_\lambda(x)$ has constant kernel dimension $m$ and is semidefinite with the same sign for all $x$. For $u\in\R^\nu$, let $H_\lambda(u)$ be the symmetric $16\times16$ matrix determined by \[ x^TH_\lambda(u)x=u^TQ_\lambda(x)u. \] Its entries are quadratic in $u$ and it is semidefinite. Consequently \[ L_u:=\{x:Q_\lambda(x)u=0\}=\ker H_\lambda(u) \] is a linear subspace. Write $d(u)=\dim L_u$; there is no parity assumption on this dimension. Choose independent $u_1,\ldots,u_m$ maximizing $\sum_i d(u_i)$ and put $S=\Span(u_1,\ldots,u_m)$. If $u\notin S$, then for every $i$ the vectors $u$ and $u_j$ ($j\ne i$) are independent, so maximality gives $d(u)\le d(u_i)$. Any $m+1$ independent coefficient vectors have $L$-intersection zero, since otherwise $\ker Q_\lambda(x)$ would have dimension at least $m+1$ at some nonzero $x$. The subspace intersection inequality gives \begin{equation}\label{eq:unrounded} d(u)\le h:=\left\lfloor\frac{16m}{m+1}\right\rfloor \qquad(u\notin S). \end{equation} The incidence space \[ \mathcal I=\{(x,u):x\ne0,\ Q_\lambda(x)u=0\} \] is a connected real-analytic rank-$m$ vector bundle over $\R^{16}\setminus\{0\}$, of dimension $16+m$. Its subset $\mathcal I_{\rm out}$ with $u\notin S$ is open and has semialgebraic dimension at most $\nu+h$. If $\nu+h<16+m$, it is empty; then every $\ker Q_\lambda(x)=S$. The scalar case $m=\nu$ is immediate. For $\nu\le8$, the only remaining nonscalar pairs are \[ \begin{array}{c|rr} (\nu,m)&h&\nu+h\\ \hline (7,3)&12&19\\ (8,2)&10&18\\ (8,3)&12&20\\ (8,4)&12&20\\ (8,5)&13&21 . \end{array} \] Suppose $\mathcal I_{\rm out}\ne\varnothing$ in one of these cases. Set $\ell=16+m-\nu$. The coefficient strata with $d(u)\le\ell-1$ have incidence dimension at most $\nu+\ell-1=15+m$. Hence the locus with $d(u)\ge\ell$ has full dimension in $\mathcal I_{\rm out}$ and contains a relatively open subset of $\mathcal I$. All $(17-\ell)$-minors of $H_\lambda(u)$ vanish there. Their pullbacks to $\mathcal I$ are real analytic. Connectedness and analytic continuation imply that they vanish on all of $\mathcal I$. Thus every \emph{nonempty} coefficient fibre has $d(u)\ge\ell$. Normalize both variables: \[ \mathcal E=\{(x,u)\in S^{15}\times S^{\nu-1}:Q_\lambda(x)u=0\}, \qquad Z=\{u\in S^{\nu-1}:\det H_\lambda(u)=0\}. \] The first projection is an $S^{m-1}$-bundle, so $\dim\mathcal E=14+m$. The second projection is a proper surjection onto the compact real algebraic set $Z$, with fibres $S(L_u)$ of dimension at least $\ell-1$. Outside $S$, their dimension is at most $h-1$ by \eqref{eq:unrounded}. Since $\mathcal E_{\rm out}$ is nonempty and open, semialgebraic dimension inequalities give \[ 15+m-h\ \le\ z:=\dim Z\ \le\ 15+m-\ell=\nu-1 . \] The relevant values are \[ \begin{array}{c|rcc} (\nu,m)&\ell&\text{possible }z&\text{minimum sphere-fibre dimension}\\ \hline (7,3)&12&6&11\\ (8,2)&10&7&9\\ (8,3)&11&6\text{ or }7&10\\ (8,4)&12&7&11\\ (8,5)&13&7&12 . \end{array} \] In every row $z<\ell-1$. The finite-degree Vietoris--Begle theorem \cite[Theorem~1.1 (Begle), p.~383]{Kryszewski}, with mod-two coefficients, gives $H^z(Z;\mathbb F_2)\cong H^z(\mathcal E;\mathbb F_2)$. The latter vanishes by the Gysin sequence of the $S^{m-1}$-bundle over $S^{15}$: the adjacent base groups in degrees $z$ and $z-m+1$ vanish in every listed case. But a nonempty compact real algebraic set of dimension $z$ has a nonzero top mod-two homology class \cite[Theorem 1.1, p.~22]{BochnakKucharz}, hence nonzero $H^z(Z;\mathbb F_2)$, a contradiction. All spaces here are compact semialgebraic and triangulable, so \v Cech and singular cohomology agree; no constant rank of the second projection is assumed. We conclude that $\ker Q_\lambda(x)=S$ for all $x\ne0$. At $x_0$ this is exactly the $\lambda$-eigenspace of $\Lambda$. Therefore for $u\in S$, $\Phi(x)u=\lambda q\Lambda^{-1}u=qu$. Symmetry removes the cross blocks. Restricting $\Phi,\Lambda$ to the Euclidean orthogonal complement preserves all hypotheses, including normalization at $x_0$ and the remaining spectrum. Induction on $\nu$ proves $\Phi=qI$. \end{proof} \subsection{The global linear factor} \begin{proof}[Proof of Proposition~\ref{prop:gramconditional}(1)] A constant coefficient change normalizes $M_{x_0}^TM_{x_0}=I$ at one unit vector while replacing $\Lambda$ by its congruent nonzero spectral matrix. For a global linear factor, $\Phi(x)=M_x^TM_x$ is polynomial homogeneous quadratic and positive definite. The nonzero spectrum of $\Jac_x$ is that of $\Lambda\Phi(x)$, so Lemma~\ref{lem:gramrigidity} gives $M_x^TM_x=qI$. For a coefficient basis vector $e_i$, write $J_ix=M_xe_i$. The identity $\Jac_xx=0$, injectivity of $M_x$, and invertibility of $\Lambda$ imply $M_x^Tx=0$, so each $J_i$ is skew. The Gram identity makes the $J_i$ orthogonal and gives $J_iJ_j+J_jJ_i=-2\delta_{ij}I$ by polarization. Diagonalizing $\Lambda$ gives the Clifford Jacobi formula. The zero-column case is simply constant curvature. \end{proof} \subsection{The rational partner and disjoint pole supports} \begin{proof}[Proof of Proposition~\ref{prop:gramconditional}(2)] On the real regular locus, $\Phi=M^TM$ is positive definite. The determinant identity $\det\Phi(z)=q(z)^\nu$ follows from the nonzero spectrum. If $\Phi$ is polynomial, continuity makes it positive semidefinite everywhere, and the determinant makes it positive definite for every nonzero real $z$. By \eqref{eq:shear} and homogeneity, $M_{X+tY}=M_X+tM_Y$ for generic $Y\perp X$ and nonzero $t$. Polynomial continuity at $t=0$ gives $\Phi(X)=I$. Lemma~\ref{lem:gramrigidity} therefore applies and yields \begin{equation}\label{eq:normalizedGram} \Phi(z)=q(z)I. \end{equation} For a nonzero root $\lambda_j$, let $I_j$ be the constant coefficient projector of $\Lambda$ and $\pi_j(z)$ the corresponding reduced Jacobi projector. From \eqref{eq:normalizedGram}, \begin{equation}\label{eq:projectorpartner} C_j(z):=M_zI_jM_z^T=q(z)\pi_j(z) \end{equation} on the regular locus, hence as a rational identity. In coordinates $z=sX+Y$, the left side has a denominator dividing a power of a polynomial $D(Y)$ independent of $s$, by \eqref{eq:shear}. The spectral-projector formula for the right side has possible denominators only powers of $q(z)=s^2+\norm Y^2$. This polynomial is irreducible over $\mathbb C$ in sixteen variables and is coprime to every nonzero $D(Y)$. Unique factorization in the polynomial ring therefore shows that each entry of $C_j$ has no denominator. Homogeneity makes it quadratic. The reduced zero-root family is the polynomial matrix \[ C_0(z)=q(z)I-zz^T-\sum_jC_j(z). \] Each $C_j$ is symmetric quadratic and satisfies $C_j(z)z=0$. The polarization argument of Section~2 (after \eqref{eq:Aproperties}) gives its sectional symmetry $\ip{C_j(x)y}y=\ip{C_j(y)x}x$. Thus every root satisfies the proportionality identity and $R$ is Jacobi-proportional. \classcor{} classifies it as Clifford or Cayley. If $\Lambda$ is nonscalar, the shifted reduced zero root and at least two distinct nonzero roots give at least three reduced eigenvalues, excluding the Cayley affine family. \end{proof} \subsection{Generic partners and the full null-cone field} \begin{lemma}\label{lem:rationalpartner} For the shifted tensor above, with $\nu\le8$, one can choose a real unit anchor $X$ and a rational homogeneous partner $M_Y$ on $X^\perp$ such that \[ \Jac_{sX+Y}=(sM_X+M_Y)\Lambda(sM_X+M_Y)^T, \] where $M_X$ is an orthonormal eigenframe. No global linearity in $Y$ is asserted. \end{lemma} \begin{proof} The generic-pair and unique two-plane factorization results \cite[Section~4, Lemma~2(1) and Lemma~3, pp.~7--9 of the preprint]{Nik2005} apply when $n\ge2\nu$, including $n=16$, $\nu=8$. Fix a suitable $X$ and its orthonormal eigenframe $M_X$. For generic $Y\perp X$, their conclusion supplies a unique partner $M_Y$. Write $M_Y=M_XF(Y)+K_XP(Y)$, where $K_X$ is an orthonormal frame of $\ker\Jac_X$, and write $2\mathcal B(X,Y)=\Jac_{X+Y}-\Jac_X-\Jac_Y$. Comparison of the mixed coefficient gives \[ P(Y)=K_X^T\,2\mathcal B(X,Y)M_X\Lambda^{-1}. \] This is linear. On the generic-pair locus $P$ has full column rank, since the images of $\Jac_X$ and $\Jac_Y$ are disjoint. The lower-upper block of $\Jac_Y$ gives \[ P\Lambda F^T=K_X^T\Jac_YM_X. \] Solving with the rational left inverse $(P^TP)^{-1}P^T$ shows that $F$, and hence $M_Y$, is rational homogeneous of degree one. The two-plane identities hold rationally by continuation from the generic real set. Defining $M_{sX+Y}=sM_X+M_Y$ gives the required coherent factor. \end{proof} \begin{proof}[Proof of Theorem~\ref{thm:extreme}] Shift the extreme root of multiplicity seven or eight to zero. Then $\nu=7$ or $8$, and $\Lambda$ is definite. Use Lemma~\ref{lem:rationalpartner} and coordinates $Y=(y_1,\ldots,y_{15})$ on $X^\perp$. Put \[ K=\R(y_1,\ldots,y_{15}),\qquad d=\sum_{i=1}^{15}y_i^2,\qquad G(s)=s^2I+sS+\Phi, \] where $S=M_X^TM_Y+M_Y^TM_X$ and $\Phi=M_Y^TM_Y$ have entries in $K$. Equality of the nonzero Jacobi spectra gives \begin{equation}\label{eq:spectralpencil} \det(tI-\Lambda G(s)) =\prod_{i=1}^{\nu}\bigl(t-(s^2+d)\mu_i\bigr) \quad\text{in }K[s,t], \end{equation} where $\mu_i$ are the nonzero roots with multiplicities. Let $L=K(\alpha)$ with $\alpha^2=-d$. The formally real field $K$ does not contain a square root of $-d$, so $[L:K]=2$. The field $L$ is the fraction field of \[ \R[s,y_1,\ldots,y_{15}]/(s^2+y_1^2+\cdots+y_{15}^2). \] By \cite[Lemma 3.11 and its proof, p.~398]{Nik2006}, a sum of at most eight squares in this field vanishes only when every term vanishes. This also follows directly by clearing denominators and applying the polynomial form of that lemma. Every constant definite real quadratic form of dimension at most eight is therefore anisotropic over $L$, after real diagonalization and rescaling. Evaluate \eqref{eq:spectralpencil} at $s=\alpha$. The matrix $N=\Lambda G(\alpha)$ has characteristic polynomial $t^\nu$, so it is nilpotent. It is self-adjoint for the symmetric bilinear form $H=\Lambda^{-1}$, since $HN=G(\alpha)$ is symmetric. This form is anisotropic over $L$ by definiteness. A self-adjoint nilpotent on an anisotropic space is zero: if its nilpotency index were $k\ge2$, choose $v$ with $w=N^{k-1}v\ne0$. Then \[ \ip ww_H=\ip v{N^{2k-2}v}_H=0, \] contradicting anisotropy. Hence $N=0$ and, since $\Lambda$ is invertible, \[ \Phi-dI+\alpha S=0. \] The $K$-linear independence of $1,\alpha$ gives separately $S=0$ and $\Phi=dI$. Thus the full Gram is $G(s)=(s^2+d)I$, polynomial on all of $\R^{16}$. Proposition~\ref{prop:gramconditional}(2) applies. There are at least three reduced roots, so its Cayley alternative is excluded. The original tensor is Clifford. \end{proof} \begin{remark} Radial Gram does not make a rational frame linear; the scalar Cayley case remains consistent with the argument. The theorem settles the extreme spectral placements of the large root. When that root lies strictly between others, $\Lambda$ is indefinite and the anisotropy step is unavailable. \end{remark} \begin{remark}[An intrinsic null-cone criterion]\label{rem:nullcriterion} Let $R$ be an Osserman algebraic curvature tensor on $\R^{16}$ with at least three distinct reduced Jacobi eigenvalues and an eigenvalue of multiplicity seven or eight. Extend its Jacobi operator $R_z=R(\,\cdot\,,z)z$ complex bilinearly, and put $q(z)=z^Tz$. Then $R$ is Clifford if and only if \[ R_z^2=0\qquad\text{for every }z\in\mathbb C^{16} \text{ with }q(z)=0. \] Equivalently, every entry of the polynomial matrix $R_z^2$ is divisible by $q$. This equivalence uses irreducibility of the complex quadratic polynomial $q$. Here is the reduction to the preceding tools. A constant-curvature shift $J_x=R_x-\lambda\Pi(x)$ satisfies \[ R_x^2=J_x^2+q(x)\bigl(2\lambda J_x+\lambda^2\Pi(x)\bigr), \] because $J_xx=0$ and $\Pi(x)^2=q(x)\Pi(x)$. Thus the null-cone condition is unchanged by the shift. Shift the large root to zero and use Lemma~\ref{lem:rationalpartner}, with the fields $K,L$ and matrices $G(s),S,\Phi$ from the proof of Theorem~\ref{thm:extreme}; no definiteness is needed here. Over $L$, set $A=M_{\alpha X+Y}$. The component $K_XP(Y)$ has full column rank, so $A$ is injective and $A^T$ is surjective. Consequently \[ J_{\alpha X+Y}^2=A\Lambda G(\alpha)\Lambda A^T=0 \quad\Longleftrightarrow\quad G(\alpha)=0. \] The null-cone hypothesis therefore implies $\Phi-dI+\alpha S=0$. Independence of $1,\alpha$ over $K$ gives $S=0$ and $\Phi=dI$. The full Gram is polynomial, so Proposition~\ref{prop:gramconditional}(2) proves the Clifford conclusion; the assumption of at least three roots excludes its Cayley alternative. Conversely, in the Clifford Jacobi formula the vectors $z,J_i z$ are mutually orthogonal and null whenever $q(z)=0$, since $\langle J_i z,J_j z\rangle=\delta_{ij}q(z)$ and $\langle z,J_i z\rangle=0$. All products of its rank-one terms then vanish, giving $R_z^2=0$. The criterion reduces the remaining indefinite cases to a polynomial divisibility statement; it does not establish that statement for general Osserman tensors. The first differentiated spectral identity does not supply a new constraint beyond constant rank. For example, with three distinct shifted roots $0,a,c$, one has $J_z^3=0$ on the null cone and \[ J_z^2B(z,w)J_z^2=0, \qquad 2B(z,w)=J_{z+w}-J_z-J_w. \] At a generic null point $z$, the partner factorization gives $\operatorname{rank}J_z=\nu$. All $(\nu+1)$-minors of $J_x$ vanish identically. Differentiating these minors gives $\langle B(z,w)t,t'\rangle=0$ for every $w$ and $t,t'\in\ker J_z$. This tangent-space statement does not require the kernel to be nondegenerate. Since $\operatorname{Im}J_z^2\subseteq\ker J_z$, it already implies the displayed sandwich identity. Polarized sectional symmetry also gives $B(t,t')z=0$ on the whole kernel. These identities do not prove $J_z^2=0$. \end{remark}