% Version v14: numeric alignment and coefficient spacing. \documentclass[11pt,a4paper]{article} \usepackage[T1]{fontenc} \usepackage{lmodern} \usepackage{amsmath,amssymb,amsthm,mathtools} \usepackage[a4paper,margin=27mm,headheight=14pt]{geometry} \usepackage{microtype,booktabs,array,enumitem,fancyhdr,xcolor} \usepackage[colorlinks=true,linkcolor=blue!45!black,citecolor=blue!45!black, urlcolor=blue!45!black,breaklinks=true]{hyperref} \hypersetup{pdftitle={An extremal vertex operator algebra of central charge c=64}, pdfauthor={Gerald H\"ohn}, pdfsubject={Construction of an extremal self-dual vertex operator algebra}} \pagestyle{fancy} \fancyhf{} \fancyhead[L]{\small An extremal VOA of central charge \(64\)} \fancyhead[R]{\small Gerald H\"ohn} \fancyfoot[C]{\thepage} \renewcommand{\headrulewidth}{0.3pt} \setlist{itemsep=3pt,topsep=5pt} \setlength{\emergencystretch}{2em} \numberwithin{equation}{section} \newtheorem{theorem}{Theorem}[section] \newtheorem{proposition}[theorem]{Proposition} \newtheorem{lemma}[theorem]{Lemma} \newtheorem{corollary}[theorem]{Corollary} \theoremstyle{definition} \newtheorem{definition}[theorem]{Definition} \newcommand{\F}{\mathbb F_2} \newcommand{\C}{\mathbb C} \newcommand{\Z}{\mathbb Z} \newcommand{\one}{\mathbf1} \newcommand{\Rep}{\operatorname{Rep}} \newcommand{\Irr}{\operatorname{Irr}} \newcommand{\Ind}{\operatorname{Ind}} \newcommand{\Res}{\operatorname{Res}} \newcommand{\Aut}{\operatorname{Aut}} \newcommand{\tr}{\operatorname{tr}} \newcommand{\im}{\operatorname{im}} \newcommand{\rank}{\operatorname{rank}} \newcommand{\ch}{\operatorname{ch}} \newcommand{\VecCat}{\operatorname{Vec}} \newcommand{\BW}{\mathrm{BW}} \newcommand{\bprod}{\boxtimes} \newcommand{\pair}[2]{\langle #1,#2\rangle} \newcommand{\ee}{\mathrm e} \title{An extremal vertex operator algebra\\ of central charge \(c=64\)} \author{Gerald H\"ohn\\[0.3em] \small Department of Mathematics, Kansas State University\\[0.3em] \small\href{mailto:gerald@monstrous-moonshine.de} {\texttt{gerald@monstrous-moonshine.de}}} \date{20 November 2026} \begin{document} \maketitle \begin{abstract} We construct an extremal self-dual vertex operator algebra of central charge \(64\). It is a simple-current extension of a strongly rational fixed-point algebra \(U\) in the tensor square of the standard self-dual Barnes--Wall lattice orbifold of central charge \(32\). The fusion group of \(U\) is a \(56\)-dimensional hyperbolic quadratic space over \(\mathbb F_2\), which we identify with the minuscule \(E_7(2)\)-module. Its weight-two labels are the rational points of the Freudenthal variety. In the exterior-power model, the map \(\mu(u\wedge v)=uv(u+v)\) on \(\mathbb F_{256}\) defines a maximal totally singular subspace avoiding that variety and hence gives the extremal extension. We also prove \(\Aut(U)\cong E_7(2)\) and determine the six distinct irreducible-module characters, whose five nonvacuum lowest conformal weights are \(2\), \(3\), \(4\), \(7/2\), \(9/2\). \end{abstract} \section{Introduction} We construct an extremal self-dual vertex operator algebra (VOA) of central charge \(64\). \subsection{Definitions and main results} All VOAs and their modules are over \(\C\). By a \emph{strongly rational} VOA we mean a VOA \(V\) that is simple, rational, \(C_2\)-cofinite, self-contragredient, and graded as \(V=\bigoplus_{n\geq0}V_n\) with \(V_0=\C\one\). For an ordinary or twisted module \(M\), write \(h(M)\) for its lowest \(L_0\)-eigenvalue, without the shift by \(c/24\). As in my thesis (\cite{Hoehn95}, Definition~2.1.1), a strongly rational VOA \(V\) is called \emph{self-dual} (also called \emph{holomorphic}) if, up to isomorphism, its adjoint module \(V\) is its only irreducible ordinary module. \emph{Extremality} is defined as in \cite{Hoehn95}, Definition~5.2.1: the graded character agrees with the vacuum Virasoro character through weight \(\lfloor c/24\rfloor\) or, equivalently, there are no nonvacuum Virasoro highest-weight vectors of positive weight at most \(\lfloor c/24\rfloor\). A self-dual VOA \(V\) of central charge \(64\) is extremal if and only if \[ V_1=0,\qquad V_2=\C\omega, \] where \(\omega\) is the Virasoro element of \(V\). \begin{theorem}\label{thm:U-structure} Let \(W=\widetilde V_{\BW_{32}}\) be the standard self-dual \((-1)\)-orbifold, let \(A=O_2(\Aut(W))\cong2^{27}\), and let \(\sigma\) exchange the factors of \(W\otimes W\). The fixed-point algebra \[ U=(W\otimes W)^{\Delta A\times\langle\sigma\rangle} \] is strongly rational of central charge \(64\), with \(U_1=0\) and \(U_2=\C\omega\). Its module category is pointed, with fusion quadratic space \[ (D,\,q)\cong(\F^{56},\,q^+),\qquad q([M])=2h(M)\pmod2. \] Here \(q^+\) denotes a nondegenerate hyperbolic quadratic form. The natural action on \(D\) identifies \(\Aut(U)\) with \(E_7(2)\) in its \(56\)-dimensional minuscule representation. Its five nonzero vector orbits have lowest conformal weights \(2\), \(3\), \(4\), \(7/2\), \(9/2\). The modules of weight two correspond exactly to the rational points of the Freudenthal variety, and the six module characters are given in Theorem~\ref{thm:six-characters}. \end{theorem} For \(d\in D\), write \(U_d\) for a representative of the corresponding simple \(U\)-module, with \(U_0=U\). \begin{theorem}\label{thm:main} There is a maximal totally singular subspace \(H\subset D\) containing no label of a weight-two module. The corresponding simple-current extension \[ V_H=\bigoplus_{d\in H}U_d \] is strongly rational and self-dual, with \[ c(V_H)=64,\qquad (V_H)_1=0,\qquad (V_H)_2=\C\omega. \] In particular, \(V_H\) is extremal. \end{theorem} Since \(64\equiv16\pmod{24}\), every homogeneous component \((V_H)_n\) is a conformal \(3\)-design (\cite{HoehnCD}, Theorem~3.1). The construction follows the extension principle developed in my paper (\cite{HoehnGenera}, Section~4, Theorems~4.2 and~4.3): a self-perpendicular isotropic subgroup of the fusion quadratic space gives a self-dual simple-current extension. We use the strongly rational formulation of \cite{EMS}, Theorem~4.2. The fusion-space identification uses the lowest-weight constituents of the permutation-twisted modules for \(W\otimes W\). Their mutual fusion weights recover the cubic norm of the split Albert algebra and hence the Freudenthal coordinates. For the extension, write \(E=\mathbb F_{256}\) as an eight-dimensional \(\F\)-space. In the exterior-power model \(\Lambda^2E^*\oplus\Lambda^2E\), define \[ \mu:\Lambda^2E\longrightarrow E,\qquad \mu(u\wedge v)=uv(u+v). \] The subspace \(\im\mu^*\oplus\ker\mu\) is maximal totally singular. A rank bound for its trace forms, together with the quadratic equations of the Freudenthal variety, proves that it avoids every rational rank-one vector. The complete weight-two locus gives an upper bound \(E_7(2)\) for \(\Aut(U)\). A permutation of four copies of \(V_{\BW_{16}}^+\) supplies the additional automorphism needed to attain that bound. The symmetry organizes the module characters into six functions. \paragraph{Conventions.} For a strongly rational VOA \(V\), write \(\Rep(V)\) for its modular tensor category of ordinary modules (\cite{Hua}, Theorem~4.6), \(\Irr(V)\) for the set of isomorphism classes of simple modules, and \(\bprod_V\) for the tensor product. The notation \(M^*\) denotes the contragredient module. Self-duality is equivalent to \(\Rep(V)\simeq\VecCat\) and is stronger than self-contragredience \(V\cong V^*\), which is already assumed in the definition of strong rationality. For the grading specified above, rationality together with \(C_2\)-cofiniteness is equivalent to regularity (\cite{ABD}, Theorem~4.5). We write \(Y(\omega,\,z)=\sum_{n\in\Z}L_nz^{-n-2}\). A Virasoro highest-weight vector is a nonzero homogeneous vector annihilated by every \(L_n\), \(n>0\). For references with a preprint link, statement numbers follow that version unless indicated otherwise. \subsection{Extremal VOAs and related literature} \label{sec:literature} In my thesis (\cite{Hoehn95}, Section~5.2), I discussed extremal self-dual VOAs of central charges \(8\), \(16\), \(24\), \(32\), \(40\); at that time self-duality of the indicated rank-\(32\) and rank-\(40\) orbifolds was still conjectural. I recorded that no extremal VOA of central charge \(c\ge48\) was then known. These examples were subsequently summarized in my conformal-design paper (\cite{HoehnCD}, Example~3.2). To my knowledge, Theorem~\ref{thm:main} gives the first new central-charge value for an extremal self-dual VOA since those examples. Following Witten \cite{Witten2007}, extremality was studied through modular reconstruction \cite{Manschot2007,ManschotMoore2010,MaloneyWitten2010,Ashrafi2020}, candidate characters for \(c=8m\) \cite{AvramisEtAl2007}, higher-genus identities \cite{GaiottoYin2007,Yin2008,GaberdielKellerVolpato2010}, and modular differential equations \cite{Gaberdiel2007,Gaiotto2008,GaberdielKeller2008}. Keller, Mathys and Zadeh (\cite{KellerMathysZadeh2018}, Section~5.6 and Appendix~B.2) treated \(c=64\); their extremal specialization gives \(278511\) weight-three Virasoro highest-weight vectors, in agreement with \eqref{eq:character}. In my preprint (\cite{Hoehn2008Large}, Section~4, Corollary~9 and Theorem~11), I ruled out the Monster-module structures suggested by Witten at \(c=48\), under the hypotheses stated there; see also \cite{Gaiotto2008}. Restrictions from topological modular forms \cite{LinPei2023} use additional hypotheses. Superconformal variants \cite{GaberdielEtAl2008,BenjaminEtAl2015,Harrison2016} concern a different extremality condition. Gem\"unden and Keller \cite{GemuendenKeller2019,GK} constructed non-extremal self-dual lattice orbifolds at \(c=48\) and \(c=72\). The proposed \(c=128\) construction in \cite{KellerRobertsRoberts2026} assumes the vanishing of an anomaly \(3\)-cocycle and has nonvacuum minimum \(4\); extremality would require minimum \(6\). Ji and Tian \cite{JiTian2026} construct full, nonchiral CFTs by asymmetric lattice orbifolds, with exact Virasoro-primary gaps \(3\) and \(4\) in total dimension \(h+\bar h\) and with unbounded central charge. \subsection{Extremal even unimodular lattices} Extremal VOAs were introduced as an analogue of extremal even lattices and extremal doubly-even codes. Known extremal even unimodular lattices include Nebe's rank-\(48\) example \cite{Nebe2014}, rank-\(64\) examples of Quebbemann \cite{Quebbemann1984} and Shimada \cite{Shimada2018,Shimada2022}, Nebe's rank-\(72\) example \cite{Nebe2012}, Bachoc and Nebe's two rank-\(80\) examples \cite{BachocNebe1998}, and Sakamoto's two rank-\(88\) examples (\cite{Sakamoto2026}, Theorem~1.1). A computational construction in rank \(112\) has been announced in \cite{Roiban2026}. These results do not directly give extremal VOAs in these central charges: a lattice VOA has a nonzero Heisenberg space in weight one, and its standard \((-1)\)-orbifold contains the \(n(n+1)/2\)-dimensional quadratic Heisenberg space in weight two. \paragraph{Acknowledgements.} I thank Christoph~A. Keller and Thomas Gem\"unden for discussions concerning extremal VOAs. I also thank Sven M\"oller for drawing my attention to \cite{JiTian2026} and Philipp H\"ohn for pointing out the rank-\(112\) lattice construction announced in \cite{Roiban2026}. These two works together motivated the approach of the present paper. \section{The fixed-point vertex operator algebra} \label{sec:U} Let \(W=\widetilde V_{\BW_{32}}\) be the standard self-dual \((-1)\)-orbifold of the Barnes--Wall lattice VOA. Put \[ A=O_2(\Aut(W)),\qquad V=W\otimes W,\qquad G=\Delta A\times\langle\sigma\rangle,\qquad U=V^G, \] where \(\Delta A=\{(a,\,a):a\in A\}\) and \(\sigma\) exchanges the two tensor factors. Thus \(G\cong(\Z/2)^{28}\). We first record the properties of \(W\) needed to analyze \(U\). I recorded the expected automorphism group of \(W\) in my conformal-design paper (\cite{HoehnCD}, Example~3.2), describing the group as likely to be \(2^{27}{:}E_6(2)\). Shimakura (\cite{Sh}, Theorem~2.30) subsequently proved the group shape \(2^{27}.E_6(2)\). For the detailed structural input used here we refer to Shimakura (\cite{Sh}, Proposition~1.8, (2.1), Proposition~2.11, Theorems~2.19 and~2.30): \begin{equation}\label{eq:W-input} c(W)=32,\quad W_1=0,\quad \dim W_2=139504,\quad A=O_2(\Aut W)\cong 2^{27},\quad \Aut(W)/A\cong E_6(2). \end{equation} The VOA \(W\) is strongly rational and self-dual: the lattice VOA is strongly rational, and its \((-1)\)-twisted minimum is \(32/16=2\). The three nontrivial \(V_{\BW_{32}}^+\)-modules have lowest weights \(1\), \(2\), \(5/2\), respectively. Thus the cyclic type is \(4\cdot2\equiv0\pmod2\), all nontrivial minima are positive, and the integral twisted summand is isotropic. The cyclic orbifold theorem (\cite{EMS}, Theorem~5.16) therefore applies. The quotient \(\Aut(W)/A\) acts irreducibly on \(A\) (\cite{Sh}, Proposition~2.11) and is transitive on the \(139503\) nonzero \(A\)-eigenspaces in the complement of \(\C\omega_W\), one of which is one-dimensional with a nontrivial character (\cite{Sh}, Section~2.4, Lemma~2.14 and Theorem~2.19). Let \(S\subset A^*\) be the set of nontrivial \(A\)-characters occurring in \(W_2\). Then \begin{equation}\label{eq:W2} W_2=\C\omega_W\oplus\bigoplus_{\chi\in S}\C_\chi, \qquad |S|=139503, \end{equation} where \(\C_\chi\) is a one-dimensional \(A\)-module of character \(\chi\), and \(S\) is a single \(\Aut(W)/A\)-orbit. Here and below \(2^{27}.E_6(2)\) denotes group shape; no splitting of this extension is assumed. Since \(G\) is solvable, the fixed-point theorem (\cite{CM}, Theorem~5.24 and Corollary~5.25), together with Theorem~4.5 of \cite{ABD}, shows that \(U\) is strongly rational. Its conformal vector is that of \(V\), so \(c(U)=64\), and \begin{equation}\label{eq:U-low} U_1=0,\qquad U_2=\C(\omega_W\otimes\one+\one\otimes\omega_W). \end{equation} Indeed \(\Delta A\)-invariance leaves just the two conformal vectors in \(V_2\), and \(\sigma\)-invariance leaves their sum. \section{Twisted modules and the fusion quadratic space} \label{sec:cohomology} \label{sec:weights} \subsection{The Albert module} We use the split Albert module \[ J=M_3(\F)\oplus M_3(\F)\oplus M_3(\F) \] with invariant cubic \begin{equation}\label{eq:norm} N(X,\,Y,\,Z)=\det X+\det Y+\det Z+\tr(XYZ). \end{equation} This is the three-matrix model in Cohen--Cooperstein \cite{CC}; their minus sign equals a plus sign in characteristic two. The differential \[ a^\#=dN_a\in J^* \] is always regarded as a \emph{covector}. Pairings are the ordinary coordinate pairing \(J^*\times J\to\F\). No identification of \(J\) with \(J^*\), and no dual-adjoint convention, is needed. In characteristic two the only nontrivial irreducible \(E_6(2)\)-modules of degree at most \(27\) are the dual modules with highest weights \(\lambda_1\) and \(\lambda_6\); see \cite{Lu}, Theorems~2.3 and~4.4, Appendix~A.51. The same degree bound shows that the irreducible \(27\)-dimensional \(\F\)-module \(A\) is absolutely irreducible: a nontrivial decomposition after extending scalars would have constituents of smaller nontrivial degree. After possibly composing the chosen quotient identification with the graph automorphism, we may therefore identify \(A\) with \(J\) and \(A^*\) with \(J^*\). There are three nonzero orbits, described by \begin{equation}\label{eq:ranks} \begin{array}{c|c|r} \text{rank}&\text{condition}&\text{orbit size}\\ \hline 1&a\ne0,\ a^\#=0&139503\\ 2&a^\#\ne0,\ N(a)=0&69193488\\ 3&N(a)=1&64884736 \end{array}. \end{equation} The rank-one stabilizer has shape \(2^{16}.\Omega_{10}^+(2)\), and the rank-three stabilizer is \(F_4(2)\) \cite{CC}. The dual module has the same orbit sizes. Consequently \(S\) in \eqref{eq:W2} is the dual rank-one orbit; its stabilizer is also identified in \cite{Sh}, Corollary~2.24. \subsection{Twisted conformal weights} For the unique \(a\)-twisted \(W\)-module \(W(a)\), set \(\rho_a=h(W(a))\), with \(\rho_0=0\). Put \(R=V_{\BW_{16}}^+\). Its irreducible modules \(R_s\) form a ten-dimensional plus-type quadratic fusion space \((S_{10},\,q_{10})\), whose polar form is denoted by \(b_{10}\). Their lowest weights are \begin{equation}\label{eq:R-weights} h(R_s)= \begin{cases} 0,&s=0,\\ 1,&s\ne0,\ q_{10}(s)=0,\\ 3/2,&q_{10}(s)=1. \end{cases} \end{equation} There are \(527\) nonzero singular labels and \(496\) nonsingular labels; the singular nonvacuum modules have \(16\)-dimensional weight-one spaces. The latter dimension also follows from transitivity on nonzero singular labels and \((V_{\BW_{16}}^-)_1\cong\C\otimes_{\Z}\BW_{16}\); see \cite{Sh}, Proposition~1.8. That proposition also gives rationality and \(C_2\)-cofiniteness of \(R\). Its self-contragredience follows from the nondegenerate invariant bilinear form on the lattice VOA: averaging over the involution shows that the restriction to \(R\) is nondegenerate. Thus \(R\) is strongly rational, and all its nontrivial irreducible modules have positive lowest weight. Shimakura's decomposition (\cite{Sh}, (2.1)) gives the full diagonal extension \begin{equation}\label{eq:diagonal-W} W=\bigoplus_{s\in S_{10}}R_s\otimes R_s. \end{equation} Permutation symmetries of simple-current extensions are treated more generally in \cite{LamYu2026}, Section~5. Let \(\iota\) be the literal flip on each summand in \eqref{eq:diagonal-W}. It is a VOA automorphism. Indeed, after choosing the same intertwining operator in the two identical factors, every projected component of the vertex operator is a scalar multiple of \(I_{s,t}\otimes I_{s,t}\), by the tensor-product theorem for intertwining operators (\cite{ADL}, Theorem~2.10) and fusion multiplicity one. Swapping the two tensor positions commutes with this expression. Shimakura's descriptions (2.2) and (2.4) place this flip in \(A\). For \(t\in S_{10}\), let \(\gamma_t\in A\) act on the \(s\)-summand by \((-1)^{b_{10}(s,\,t)}\). Write \[ \ch_s(z)=\tr_{R_s}\exp\bigl(2\pi i z(L_0-16/24)\bigr). \] For \(g\in\Aut(W)\), write \(Z_W(g;\,z)=\tr_W\bigl(g\exp(2\pi iz(L_0-32/24))\bigr)\). The trace function and its modular transform are \begin{align} Z_W(\iota\gamma_t;\,z) &=\sum_s(-1)^{b_{10}(s,\,t)}\ch_s(2z),\,\label{eq:flip-trace}\\ Z_W(\iota\gamma_t;\,-1/z) &=32\,\ch_t(z/2).\label{eq:flip-S} \end{align} For the second identity, use \(S_{s,t}=2^{-5}(-1)^{b_{10}(s,\,t)}\) (\cite{EMS}, Theorem~3.4) and Fourier orthogonality on the group of order \(2^{10}\). Modular covariance and its normalization (\cite{EMS}, Theorem~2.3 and Proposition~5.5) identify the right-hand side of \eqref{eq:flip-S} with the character of the unique \((\iota\gamma_t)\)-twisted \(W\)-module. Comparing leading exponents gives \[ \rho_{\iota\gamma_t}-\frac{32}{24} =\frac12\left(h(R_t)-\frac{16}{24}\right),\qquad \rho_{\iota\gamma_t}=1+\frac{h(R_t)}2. \] Thus the three possible absolute weights are \(1\), \(3/2\), \(7/4\). \begin{proposition}\label{prop:rho} The twisted lowest weights are \[ \rho_a= \begin{cases} 1,&\rank(a)=1,\\ 3/2,&\rank(a)=2,\\ 7/4,&\rank(a)=3. \end{cases} \qquad 4\rho_a\equiv N(a)\pmod2. \] For rank-one \(a\), \(\dim W(a)_1=32\). \end{proposition} \begin{proof} The traces of \(\iota\gamma_t\) on \(W_2\), for the three types of \(t\) in \eqref{eq:R-weights}, are respectively \[ 8432,\qquad 240,\qquad -272. \] The vacuum summand contributes zero flip trace in total weight two. Each nonzero singular summand contributes \(16\) times its sign. For \(t\ne0\), the singular Fourier sum is \(16(-1)^{q_{10}(t)}-1\); for \(t=0\), it is \(527\). The three distinct traces must represent the three nonzero \(E_6(2)\)-orbits in \(A\). Since \(\dim W_2^A=1\), character averaging gives \begin{equation}\label{eq:trace-average} 139504+139503\,t_1+69193488\,t_2+64884736\,t_3=2^{27}, \end{equation} where \(t_i\) is the trace on the rank-\(i\) orbit. Among the six permutations of \(8432\), \(240\), \(-272\), the unique solution is \[ (t_1,\,t_2,\,t_3)=(8432,\,240,\,-272). \] Together with \eqref{eq:flip-S}, this assigns twisted weights \(1\), \(3/2\), \(7/4\) to ranks one, two, and three, respectively. For \(t=0\), the leading coefficient of \(\ch_0\) is one, so \eqref{eq:flip-S} gives \(\dim W(\iota)_1=32\). Conjugacy gives the same dimension throughout the rank-one orbit. \end{proof} \subsection{An elementary abelian orbifold criterion} Let a finite group \(B\) act faithfully on a strongly rational self-dual VOA \(\mathcal V\), with \(\mathcal V^B\) strongly rational. The \(B\)-graded category of twisted \(\mathcal V\)-modules has one invertible simple object \(\mathcal V(b)\) in each homogeneous component (\cite{McR}, Theorems~4.15 and~4.17). Its associator defines a class \[ \mathfrak a_B(\mathcal V)\in H^3(B,\,\C^\times), \] with trivial coefficient action. We call this the \emph{orbifold \(3\)-cocycle class}; the group grading is part of the definition. Restriction to the components indexed by a subgroup gives the restriction of this class. We use the following restriction calculation for elementary abelian \(2\)-groups. \begin{lemma}\label{lem:cyclic-detection} Let \(\Gamma=(\Z/2)^r\). Restriction to all subgroups of order two gives an injective map \[ H^3(\Gamma,\,\C^\times)\longrightarrow \prod_{0\ne v\in \Gamma}H^3(\langle v\rangle,\,\C^\times). \] With the latter factors identified with \(\F\), its image consists of the functions \begin{equation}\label{eq:boolean-type} f(v)=\sum_i c_i v_i+\sum_{i0}\qquad(1\ne g\in\Gamma). \] Then \(\mathcal V^\Gamma\) is strongly rational, every nonvacuum irreducible \(\mathcal V^\Gamma\)-module has positive lowest weight, and \[ \mathfrak a_\Gamma(\mathcal V)=0,\qquad \Rep(\mathcal V^\Gamma)\simeq Z(\VecCat_\Gamma) \] as ribbon categories. In particular, its fusion group is elementary abelian of order \(|\Gamma|^2\), with hyperbolic quadratic form \([M]\mapsto2h(M)\pmod2\). \end{proposition} \begin{proof} Since \(\Gamma\) is solvable, fixed-point regularity (\cite{CM}, Theorem~5.24 and Corollary~5.25), together with Theorem~4.5 of \cite{ABD}, gives strong rationality of \(\mathcal V^\Gamma\). Every simple \(\mathcal V^\Gamma\)-module is a constituent of some \(\mathcal V(g)\), by \cite{McR}, Theorems~4.15 and~4.17. The nontrivial twisted modules have positive minima by hypothesis, and only the vacuum constituent of \(\mathcal V|_{\mathcal V^\Gamma}\) contains weight zero. This proves positivity. For every nonidentity \(g\), the cyclic type \(4h(\mathcal V(g))\pmod2\) is zero. Thus Lemma~\ref{lem:cyclic-detection} gives \(\mathfrak a_\Gamma(\mathcal V)=0\). The center-description theorem (\cite{DNR}, Theorem~6.2(4) and Remark~7.2) now applies. Positivity identifies categorical dimensions with positive Frobenius--Perron dimensions and hence identifies the spherical structure and conformal-weight ribbon twist. The simple objects of \(Z(\VecCat_\Gamma)\) are the pairs \((g,\,\chi)\), with componentwise product and twist \(\chi(g)\). This proves the assertions about the fusion group and quadratic form. \end{proof} \subsection{Application to the doubled action} Write \(d_a=(a,\,a)\) and \(p_a=(a,\,a)\sigma\). The nonidentity elements of \(G\) are \(d_a\) for \(a\ne0\) and \(p_a\) for all \(a\). Their twisted lowest weights are \begin{equation}\label{eq:V-weights} h(V(d_a))=2\rho_a,\qquad h(V(p_a))=\frac{c(W)}{16}=2. \end{equation} Indeed, \(V(d_a)=W(a)\otimes W(a)\). Also \(p_a\) is conjugate to \(\sigma\) by \((a,\,1)\), and a transposition twist satisfies \(L_0^{\mathrm{tw}}=\tfrac12L_0+c(W)/16\), where the central charge is that of one tensor factor (\cite{BDM}, proof of Theorem~3.9(3), Theorem~4.4). The lowest-weight space of each \(V(p_a)\) is one-dimensional. \begin{proposition}\label{prop:associator-zero} The orbifold \(3\)-cocycle class \(\mathfrak a_G(V)\) is zero, and \begin{equation}\label{eq:actual-category} \Rep(U)\simeq Z(\VecCat_G),\qquad D=\Irr(U)\cong\F^{56}. \end{equation} This is a ribbon equivalence. Every simple object is invertible, and every nonvacuum irreducible module has positive lowest weight. The binary conformal-weight form \begin{equation}\label{eq:q-convention} q([M])=2h(M)\pmod2,\qquad b(d,\,e)=q(d+e)+q(d)+q(e) \end{equation} is nondegenerate and hyperbolic. In particular, \(q(d)=0\) means integral conformal weight. \end{proposition} \begin{proof} Every nonidentity element of \(G\) is either \(d_a\), \(a\ne0\), or \(p_a\). Proposition~\ref{prop:rho} and \eqref{eq:V-weights} show that their twisted modules have positive half-integral minima. Proposition~\ref{prop:pointed} therefore applies to the full group \(G\). \end{proof} The \emph{twisting-label homomorphism} \(\nu\) gives the canonical exact sequence \[ 0\longrightarrow G^*\longrightarrow D \xrightarrow{\nu}G\longrightarrow0, \qquad (-1)^{b(\chi,\,d)}=\chi(\nu(d)). \] Here \(\chi\in G^*\), \(d\in D\), and \(\nu([M])=g\) when \(M\) is a constituent of \(V(g)|_U\). The kernel \(G^*\) labels the eigenspaces in \(V|_U\). A compatible identification \(D\cong G\oplus G^*\) is noncanonical; in such coordinates the twist of \((g,\,\chi)\) is \(\chi(g)\). \section{The Freudenthal module and the weight-two locus} \label{sec:geometry} \label{sec:freudenthal} The fusion category determines conformal weights modulo \(\Z\). To identify the weight-two locus, we first determine certain absolute fusion weights. \subsection{Permutation-twisted modules and absolute fusion weights} \label{sec:fusion} For \(a\in A\), let \(F_a\) be the unique simple \(U\)-module containing the one-dimensional subspace \(V(p_a)_2\) of \(V(p_a)\). Then \[ h(F_a)=2,\quad q(F_a)=0,\quad \nu(F_a)=p_a. \] The automorphism \(r_c=(c,\,1)\) normalizes \(G\) and sends \(F_a\) to \(F_{a+c}\). \begin{proposition}\label{prop:fusion-weight} For all \(a\) and \(b\) in \(A\), \begin{equation}\label{eq:exact-fusion-weight} h(F_a\bprod_U F_b)=2\rho_{a+b}. \end{equation} \end{proposition} To prove the proposition, put \[ X=W^A,\qquad Y=(X\otimes X)^\sigma=V^{\mathcal K},\qquad \mathcal K=(A\times A)\rtimes\langle\sigma\rangle . \] Then \(G\trianglelefteq\mathcal K\), \(\mathcal K/G\cong A\), and each \(r_c\) fixes \(Y\) pointwise. As in Section~\ref{sec:U}, the finite \(2\)-group orbifold theorems (\cite{CM}, Theorem~5.24 and Corollary~5.25; \cite{ABD}, Theorem~4.5) show that \(X\) and \(Y\) are strongly rational. Completeness of twisted-module constituents and the corresponding equivariantization equivalence follow from \cite{McR}, Theorems~4.15 and~4.17. Every nonvacuum \(X\)-module has positive weight: this follows from Proposition~\ref{prop:rho} for twisted constituents, and from \(W_0=\C\one\) for untwisted ones. The untwisted \(Y\)-modules arise from \(P\otimes Q\), with \(P\) and \(Q\) in \(\Irr(X)\); only the symmetric vacuum constituent contains weight zero. Permutation-twisted constituents have weights at least \(2+h(P)/2>0\). Thus every nonvacuum \(Y\)-module also has positive weight. For the full twisted module \(T_\sigma(X)\), the fusion tensor square over \(X\otimes X\) is determined by \cite{DongXuYu2025}, Theorem~4.6. We need the following statement for its lowest-weight \(Y\)-constituent. \begin{lemma}[Lowest-weight summands of a tensor square]\label{lem:lowest-weight} Let \(T\) be the simple \(Y\)-constituent of \(T_\sigma(X)\) containing its one-dimensional homogeneous subspace of weight \(c(X)/16\). Then \(T\cong T^*\), and for every \(P\in\Irr(X)\) the fusion product \(T\bprod_YT\) contains a simple summand of lowest weight \(2h(P)\), arising from \((P\otimes P^*)|_Y\). \end{lemma} \begin{proof} The category \(\Rep(Y)\) is rigid and semisimple. All modules used below restrict to finite-length ordinary \(Y\)-modules. Assign the modules \(T_\sigma(X)\), \(P\otimes P^*\), \(T_\sigma(X)\) to the marked points \(0\), \(1\), \(\infty\) of the base sphere, with monodromies \(\sigma\), \(1\), \(\sigma\). The connected covering is \(z=w^2\); the modules at the marked points \(0\), \(1\), \(-1\), \(\infty\) of the covering sphere are \(X\), \(P\), \(P^*\), \(X\). Propagation of the vacuum modules reduces the conformal block to the contragredient pairing between \(P\) and \(P^*\). Here conformal blocks and propagation are used in the algebraic-analytic definitions of \cite{Gui}, Sections~1.2 and~1.3. Choose lowest-weight vectors \(p\), \(p^*\) with \(\pair{p^*}{p}\ne0\). The cover block is nonzero on \((\one,\,p,\,p^*,\,\one)\). At the two unramified marked points the changes to lifted coordinates have nonzero derivatives; positive Virasoro modes annihilate lowest-weight vectors, so these changes only multiply the value by nonzero scalars. Propagation of the vacuum vectors also preserves its nonzero value. Gui's cover theorem (\cite{Gui}, Theorem~2.25) identifies the base and cover blocks as the \emph{same underlying linear functional}, not merely as abstract vector spaces. Therefore the base block is nonzero on \((t,\,p\otimes p^*,\,t)\), where \(t\) spans the lowest-weight space of \(T_\sigma(X)\). By \cite{Gui}, Theorem~4.10, this block determines a twisted intertwining operator. Restricting the vertex operators to \(Y\) gives an ordinary \(Y\)-intertwining operator by \cite{Gui}, Remark~4.6: the relevant \(L_0\)-eigenspaces are finite-dimensional and the spectrum is bounded below. The conformal vector is unchanged. Restrict the first and third entries to \(T\). This lowest-weight coefficient remains nonzero. Decomposing \(p\otimes p^*\) among simple \(Y\)-constituents gives a constituent \(M\) for which the three-point block with \(T\), \(T\) is nonzero. Its lowest weight is exactly \(2h(P)\): this weight occurs in \(M\), whereas the entire ambient module \(P\otimes P^*\) has no smaller weight. Contragredient compatibility (\cite{Gui}, Theorem~2.23) makes \(T_\sigma(X)\) self-contragredient. Its nondegenerate graded pairing pairs its unique lowest-weight line with itself; hence \(T^*\cong T\). The restricted three-point block therefore gives a nonzero intertwining operator of type \(\binom{T}{M\ T}\). Rigidity gives \[ 0\ne\operatorname{Hom}_Y(M\bprod_YT,\,T) \cong\operatorname{Hom}_Y(M,\,T\bprod_YT^*). \] By semisimplicity, \(M\) is a simple summand of \(T\bprod_YT\), as required. \end{proof} \begin{proof}[Proof of Proposition~\ref{prop:fusion-weight}] \emph{Step 1: the simple-current extension \(Y\subset U\).} Write \(W=\bigoplus_{\chi\in A^*}W_\chi\) as an \(X\)-module. Its summands are simple currents indexed by \(A^*\), and \begin{equation}\label{eq:U-over-Y} U=\bigoplus_{\chi\in A^*}(W_\chi\otimes W_\chi)^+ \end{equation} is a simple-current extension of \(Y\) with \(|A|\) summands. Indeed, \(\mathcal K/G\cong A\) acts on \(U\), its fixed algebra is \(Y\), and \((c,\,1)\) acts on the \(\chi\)-summand by \(\chi(c)\). The tensor subcategory generated by these summands is \(\Rep(A)\) (\cite{DNR}, Theorem~6.1). Consequently each summand in \eqref{eq:U-over-Y} is invertible in \(\Rep(Y)\). Moreover \(U\), as an algebra object in this subcategory, corresponds to \(\operatorname{Fun}(A)\) (\cite{DNR}, Theorem~6.3). In particular the multiplication of any two character summands is nonzero. \emph{Step 2: the restriction of \(F_a\).} In the permutation-twisted module \(V(\sigma)=T_\sigma(W)\), the underlying space is \(W\), with grading \(L_0^W/2+2\). The construction (\cite{BDM}, (2.7), (3.13) and Theorem~3.9) lifts \(d_a\) to \(a\) and \(\sigma\) to \((-1)^{L_0^W}\); these commuting operators fix the lowest-weight line. By the equivariantization above, their common fixed space is \(F_0=\bigoplus_{m\ge0}X_{2m}\). Restriction of the same construction to \(X\otimes X\) identifies this space, as a \(Y\)-module, with the constituent \(T\) of \(T_\sigma(X)\) containing its lowest-weight line. Since \(r_a\) fixes \(Y\) pointwise and sends \(F_0\) to \(F_a\), \begin{equation}\label{eq:restriction} \Res_Y^U F_a\cong T\qquad(a\in A). \end{equation} \emph{Step 3: local induction.} Every simple-current summand \(K\) of \(U\) acts nontrivially on the simple underlying object \(T=\Res F_0\): otherwise the module axiom and the nonzero product \(K^{-1}\bprod K\to\one\) would contradict the unit action. Thus \(K\bprod T\cong T\). Since \(\theta_K=1\), balancing shows that \(T\) centralizes \(U\), as do \(T\bprod_YT\) and all its simple summands. Free induction is consequently local and braided monoidal (\cite{CKM}, Proposition~2.65 and Theorem~2.67). The local-module tensor product agrees with VOA fusion by \cite{CKM}, Theorem~3.65, since \(Y\) and \(U\) are strongly rational and \(U\) is a finite object of \(\Rep(Y)\). By induction--restriction adjunction (\cite{CKM}, Lemma~2.61) and \eqref{eq:restriction}, each of the \(|A|\) distinct modules \(F_a\) occurs once in \(\Ind_Y^U T\). They exhaust it because \(\Res\Ind T=U\bprod_YT\cong T^{\oplus|A|}\). Monoidality now gives \[ \Ind_Y^U T\cong\bigoplus_{a\in A}F_a,\qquad \Ind(T\bprod_YT)\cong \left(\bigoplus_aF_a\right)\bprod_U \left(\bigoplus_bF_b\right). \] \emph{Step 4: the absolute conformal weight.} Fix \(u\ne0\), and choose an \(X\)-constituent \(P\) of \(W(u)\) containing a lowest-weight vector. Then \(h(P)=\rho_u\), and \(P^*\) also occurs in \(W(u)|_X\), since \(u^{-1}=u\). Lemma~\ref{lem:lowest-weight} gives a constituent \(M\) of \(T\bprod_YT\), also arising in \(W(u)\otimes W(u)\), with \(h(M)=2\rho_u\). The unit summand of \(U\) gives an injective \(Y\)-map \(M\to\Res\Ind M\); thus some simple \(U\)-summand \(N\) of \(\Ind M\) contains a vector of weight \(2\rho_u\). We claim that \(\nu(N)=d_u\). Using \(Y=V^{\mathcal K}\), put \(g=\nu(N)\). Then \[ \operatorname{Hom}_U(\Ind M,\,N) \cong\operatorname{Hom}_Y(M,\,\Res N)\ne0. \] Thus \(M\) occurs in both \(V(g)|_Y\) and \(V(d_u)|_Y\). Twisted modules labeled by different \(\mathcal K\)-conjugacy classes have disjoint simple \(Y\)-constituents (\cite{DNR}, Theorem~5.1(1)). Since \(d_u\) is central in \(\mathcal K\), its conjugacy class is a singleton, so \(g=d_u\). The displayed induced product decomposition consequently makes \(N\) one of \(F_a\bprod F_{a+u}\). Every module with twisting label \(d_u\) occurs in \(W(u)\otimes W(u)\), whose minimum is \(2\rho_u\). Hence \(h(N)=2\rho_u\). The normalizer translations act transitively on the pairs \((a,\,a+u)\) and preserve conformal weight, proving the equality for every pair. For \(u=0\), the equality follows from \(F_a^2=U\), since \(D\) has exponent two. \end{proof} \subsection{The Freudenthal coordinates} Define \[ \mathcal E=\F\oplus J\oplus\F\oplus J^*,\qquad Q(\alpha,\,a;\,\beta,\,B)=\alpha\beta+\pair{B}{a}. \] Its polar form is denoted by \(B_{\mathcal E}\). For \(a\in J\), put \begin{equation}\label{eq:f-vector} f_a=(1,\,a;\,N(a),\,a^\#). \end{equation} Euler's identity for a cubic in characteristic two gives \(\pair{a^\#}{a}=N(a)\). Cubic expansion then gives \begin{equation}\label{eq:gram} Q(f_a)=0,\qquad B_{\mathcal E}(f_a,\,f_b)=N(a+b). \end{equation} On the fusion group, Propositions~\ref{prop:rho} and \ref{prop:fusion-weight} give \[ b(F_a,\,F_b)=4\rho_{a+b}\pmod2=N(a+b). \] \begin{proposition}\label{prop:coordinates} There is a unique linear isometry \(\Phi:(D,\,q)\to(\mathcal E,\,Q)\) sending \(F_a\) to \(f_a\). For the character \begin{equation}\label{eq:character-module} E_{\beta,B}:\quad d_a\sigma^s\longmapsto (-1)^{\beta s+\pair{B}{a}}, \end{equation} its corresponding simple module has image \((0,\,0;\,\beta,\,B)\). \end{proposition} \begin{proof} The vectors \(f_a\) span \(\mathcal E\). Indeed, a linear functional vanishing on all of them would give an identically zero Boolean polynomial \[ c+\ell(a)+\pair{a^\#}{v}+dN(a). \] In the coordinates \eqref{eq:norm}, \(N\) is a sum of \(45\) distinct squarefree cubic monomials. Each of its \(27\) derivatives has five quadratic monomials, and these \(135\) monomials are mutually distinct. The derivative polynomials are therefore independent. Constant, linear, quadratic, and cubic squarefree monomials are independent as functions on \(\F^{27}\), so all coefficients vanish. Because the form on \(\mathcal E\) is nondegenerate, the Gram matrix of the \(f_a\) has rank \(56\). The equal Gram matrix of the \(F_a\) forces them to span the \(56\)-dimensional space \(D\). By equality of the Gram matrices, a relation among one family gives a linear combination of the other that is orthogonal to every member of that family. Nondegeneracy forces this combination to vanish. Thus the assignment extends to an isometry of polar forms. Both quadratic forms vanish on the spanning family; their difference is linear and hence zero. This proves existence and uniqueness. The module indexed by \eqref{eq:character-module} pairs with \(F_a\) as \(\beta+\pair{B}{a}\). The vector \((0,\,0;\,\beta,\,B)\) has exactly these pairings with every \(f_a\). Nondegeneracy proves the final assertion. \end{proof} \subsection{The weight-two locus} For rank-one \(u\), put \[ \beta_u(a)=\pair{a^\#}{u},\qquad H_u(a)=(a+u)^\#+a^\#+u^\#. \] Since \(u^\#=0\), cubic expansion yields \begin{equation}\label{eq:boundary-points} f_a+f_{a+u}=(0,\,u;\,\beta_u(a),\,H_u a). \end{equation} \begin{lemma}\label{lem:rank10} For every nonzero rank-one \(u\), the linear map \(H_u\) has rank \(10\), and \(\beta_u\) is constant on its fibers. Consequently \eqref{eq:boundary-points} gives exactly \(2^{10}\) distinct vectors as \(a\) varies. \end{lemma} \begin{proof} The \(E_6(2)\)-action preserves \(N\) and is transitive on rank-one vectors. It suffices to take \(u=(E_{11},\,0,\,0)\) in \eqref{eq:norm}. Direct differentiation gives \begin{equation}\label{eq:beta-normal} \beta_u(X,\,Y,\,Z) =x_{22}x_{33}+x_{23}x_{32} +y_{11}z_{11}+y_{12}z_{21}+y_{13}z_{31}. \end{equation} This consists of five disjoint hyperbolic pairs; its remaining seventeen coordinates form a radical on which it is identically zero. Its polar map is \(H_u\), proving the claims for this \(u\). Norm invariance gives \[ \beta_{gu}(ga)=\beta_u(a),\qquad H_{gu}(ga)=(g^{-1})^*H_u(a), \] which transports the assertions to every rank-one vector. \end{proof} By Proposition~\ref{prop:fusion-weight}, all \(1024\) corresponding distinct \(U\)-modules have weight two. The full \(d_u\)-twisted module has a lowest-weight space of dimension \[ \dim\bigl(W(u)_1\otimes W(u)_1\bigr)=32^2=1024. \] Each of these distinct modules contributes at least one line to this space. Equality proves that they exhaust it and that each has a one-dimensional lowest-weight space. \begin{proposition}\label{prop:bad-set} Every nonvacuum simple \(U\)-module has lowest weight at least two. The images under \(\Phi\) of the labels of weight-two modules form the disjoint union \begin{align} \mathcal B_{\mathrm{unt}} &=\{(0,\,0;\,1,\,0)\}\cup \{(0,\,0;\,\beta,\,B):B\in S,\ \beta\in\F\},\,\label{eq:bad-unt}\\ \mathcal B_{\mathrm{diag}} &=\{(0,\,u;\,\beta_u(a),\,H_u a): u\ne0,\ u^\#=0,\ a\in J\},\,\label{eq:bad-diag}\\ \mathcal B_{\mathrm{perm}} &=\{f_a:a\in J\}.\label{eq:bad-perm} \end{align} Put \(\mathcal B=\mathcal B_{\mathrm{unt}}\sqcup \mathcal B_{\mathrm{diag}}\sqcup\mathcal B_{\mathrm{perm}}\). The three sets have respective sizes \[ 279007,\qquad 142851072,\qquad 134217728, \] and their total size is \(277347807\). \end{proposition} \begin{proof} The ordinary module \(V\) has \(V_2=W_2\otimes\one+\one\otimes W_2\). Its \(\sigma\)-eigenspaces and \eqref{eq:W2} give \eqref{eq:bad-unt}, with the vacuum constituent omitted. The preceding dimension argument gives exactly \eqref{eq:bad-diag} for rank-one diagonal twisting labels. The other diagonal minima are \(3\) and \(7/2\). Every \(V(p_a)\) has a one-dimensional weight-two subspace, giving \eqref{eq:bad-perm}. The ordinary module \(V\) has no weight-one vectors, and every nontrivially twisted module has minimum at least two. Completeness of their irreducible \(U\)-constituents excludes additional cases. \end{proof} We call \(\mathcal B\) the \emph{obstruction set} and also view it as a subset of \(D\) via \(\Phi^{-1}\). \subsection{The Freudenthal variety} Put \(k=\overline{\F}\). Let \begin{equation}\label{eq:freudenthal-chart} \Omega=\overline{\{[1,\,a;\,N(a),\,a^\#]:a\in J\otimes_{\F} k\}} \ \subset\ \mathbb P(\mathcal E\otimes_{\F} k), \end{equation} where the bar denotes Zariski closure over \(k\). The cubic \eqref{eq:norm} is the norm of the split Albert algebra: it is the first Tits construction for \(M_3(\F)\) with parameter \(1\); see \cite{GPR}, (12.3) and Example~12.2(b). The associated cubic embedding \eqref{eq:freudenthal-chart} is the standard open chart of the Freudenthal variety \(E_7/P_7\) in the minuscule \(56\)-dimensional representation. For the cubic construction, its equivalence with the weight-coordinate model, and the small-field convention, see \cite{DSSV}, Definition~10.1, Theorem~10.4, Remark~10.5 and Section~10.4. The associated split group of type \(E_7\) over arbitrary base rings is described in \cite{GPR}, Sections~16--17, especially Theorem~17.4. To compare with the vector-valued Jordan adjoint, write \(a^{\#_J}\in J\) for that adjoint and \(T_J\) for the nondegenerate trace pairing of the Albert algebra. The identity \[ T_J(a^{\#_J},\,-)=dN_a \] is \cite{GPR}, (6.4), valid in all scalar extensions. Transporting the cited construction through this trace pairing therefore gives exactly our covector convention \(a^\#=dN_a\). In particular, \(\Omega\) is defined over \(\F\). A nonzero vector whose projective class belongs to \(\Omega\) will be called a \emph{Freudenthal rank-one vector}. \begin{proposition}\label{prop:bad-in-freudenthal} The obstruction set \(\mathcal B\) of Proposition~\ref{prop:bad-set} is exactly the set of rational Freudenthal rank-one vectors. Equivalently, identifying a rational projective point with its unique nonzero \(\F\)-representative, \(\mathcal B=\Omega(\F)\). \end{proposition} \begin{proof} The vectors in \(\mathcal B_{\mathrm{perm}}\) belong to the chart by definition. Let \(0\ne u\in J\) have Albert rank one. Then \(u^\#=0\), and Euler's identity gives \(N(u)=0\). For an indeterminate \(t\), cubic expansion gives the polynomial identities \[ \begin{aligned} N(a+tu)&=N(a)+t\,\pair{a^\#}{u},\\ (a+tu)^\#&=a^\#+tH_u a. \end{aligned} \] Consequently \[ f_{a+tu}=f_a+t(0,\,u;\,\beta_u(a),\,H_u a). \] The point at infinity of this projective line is \([0,\,u;\,\beta_u(a),\,H_u a]\). It belongs to \(\Omega\), proving the assertion for \(\mathcal B_{\mathrm{diag}}\). Let \(B_0\in J^*\) be the coordinate covector extracting \(x_{33}\). It is a weight vector of the dual minuscule \(E_6\)-module and belongs to its rank-one orbit \(S\); compare the three-matrix model and the duality in \cite{CC}. For \(\varepsilon\in\F\), take \[ a(t)=(\operatorname{diag}(t,\,t,\,\varepsilon),\,0,\,0). \] Then \[ N(a(t))=\varepsilon t^2,\qquad a(t)^\#=(\operatorname{diag}(\varepsilon t,\,\varepsilon t,\,t^2),\,0,\,0), \] where the second expression lists covector coordinates in the dual coordinate basis. The leading term of \(f_{a(t)}\), after multiplication by \(t^{-2}\), gives the point \([0,\,0;\,\varepsilon,\,B_0]\) at infinity. The \(E_6(2)\)-action \[ (\alpha,\,a;\,\beta,\,B)\longmapsto (\alpha,\,ga;\,\beta,\,(g^{-1})^*B) \] preserves the chart and is transitive on \(S\). Hence all the points \([0,\,0;\,\varepsilon,\,B]\), \(B\in S\), lie in \(\Omega\). Finally, the curve \(a(t)=(tI_3,\,0,\,0)\), scaled by \(t^{-3}\), gives \([0,\,0;\,1,\,0]\). Thus \(\mathcal B\subset\Omega(\F)\). The transitive \(E_7(2)\)-action has point stabilizer \(P_7(2)\), of order \(2^{27}|E_6(2)|\). The split-group order formulas give \[ |\Omega(\F)| =\frac{|E_7(2)|}{2^{27}|E_6(2)|} =(2^5+1)(2^9+1)(2^{14}-1) =277347807. \] This equals \(|\mathcal B|\) by Proposition~\ref{prop:bad-set}, so the containment is an equality. \end{proof} \section{An extremal self-dual extension} \label{sec:extension} \label{sec:theoretical} We now construct one maximal totally singular subspace avoiding \(\mathcal B\). All varieties and subspaces in this section are defined over \(\F\), unless otherwise stated. \subsection{The extension criterion} \begin{corollary}[Extremality criterion]\label{cor:extremality} For any maximal totally singular subspace \(H\subset D\), the simple-current extension \[ V_H=\bigoplus_{d\in H}U_d \] is strongly rational and self-dual. It is extremal if and only if \(H\cap\mathcal B=\varnothing\). \end{corollary} \begin{proof} Following \cite{HoehnGenera}, Section~4, we use the self-dual extension associated with \(H\). The strongly rational existence theorem (\cite{EMS}, Theorem~4.2), the VOA form of the Lagrangian-subgroup construction (\cite{DMNO}, Section~5.3), applies: \(U\) is strongly rational, its nonvacuum irreducible modules have positive lowest weight, and \(H\) is isotropic and self-perpendicular. Every included module has integral lowest weight. By \eqref{eq:U-low}, the vacuum summand contributes no weight one and only the conformal vector in weight two. Proposition~\ref{prop:bad-set} lists exactly the nonvacuum integral-weight summands which contribute in these degrees. Thus \(V_H\) is extremal exactly when none is included. \end{proof} \subsection{The exterior-power model} Let \(E=\mathbb F_{256}\), viewed for the moment simply as an eight-dimensional \(\F\)-space. Put \begin{equation}\label{eq:exterior-model} \mathcal M=\Lambda^2E^*\oplus\Lambda^2E,\qquad Q_8(X,\,Y)=\pair{X}{Y}. \end{equation} The pairing in \eqref{eq:exterior-model} is the natural perfect pairing between the two exterior powers. In a basis, \(X\) and \(Y\) are represented by alternating \(8\times8\) matrices, and \[ Q_8(X,\,Y)=\sum_{0\le i