\section{The rank-one obstruction}\label{sec:rank} We prove an upper bound for every knot obtained by following \(U\) with a braid in the subgroup acting on \eqref{eq:framed}. That subgroup fixes the first and fourth meridian words. These two fixed positions suffice for the rank-one argument. \begin{lemma}\label{lem:spans} Suppose that \(\tau\) belongs to \(\langle\sigma_2,g,\sigma_5,\ldots,\sigma_{b-1}\rangle\) and \(\cl{U\tau}\) is a knot. Let its group act on a finite-dimensional space \(V\) over a field so that each meridian has displacement of rank one. Write each meridian as \(M=I-u f\), where \(u\in V\) and \(f\in V^*\). Then at least one of the spans of all the \(u\)'s and all the \(f\)'s has dimension at most \(b-1\), and both dimensions are at most \(b\). \end{lemma} \begin{proof} Each of the two spans is already spanned by its \(b\) top-strand vectors or covectors. The root span of the generating meridians is invariant under those generators and their inverses, and the covector span is invariant under right multiplication by the same matrices. Conjugating a generator therefore introduces neither a new root direction nor a new covector direction. Suppose both spans have dimension \(b\). In particular both families of four vectors and covectors on the first four strands are separately independent. Set \[ \mu=\det M\ne0,\qquad f(u)=1-\mu,\qquad M^{-1}=I+\mu^{-1}u f. \] The first two identities hold for every meridian, with one common \(\mu\). When \(M_a=I-u_a f_a\), conjugation by a matrix \(J\) replaces its pair by \((Ju_a,f_aJ^{-1})\). Two pairs for the same nonzero displacement differ by reciprocal scaling. Here is the crossing trace through \(U\). We use arc subscripts on the matrices \(M_i\) in this proof. A row \((o,a,c,\epsilon)\) represents \(M_c=M_o^\epsilon M_aM_o^{-\epsilon}\). The input labels are \((1,0,8,13)\) and the output labels are \((1,6,16,13)\). \begin{equation}\label{eq:prefix} \begin{array}{c|rrrr@{\qquad}c|rrrr} 0&0&1&2&-&1&8&2&3&-\\ 2&3&13&14&+&3&14&3&4&+\\ 4&4&8&9&-&5&9&14&15&+\\ 6&15&9&10&+&7&10&4&5&-\\ 8&10&0&1&-&9&1&10&11&-\\ 10&5&11&12&-&11&12&5&6&-\\ 12&6&12&13&-&13&13&15&16&+ \end{array} \end{equation} The repeated outer labels express closure, since \(\tau\) fixes the first and fourth entries. Each row is obtained by \(T\) or \(T^{-1}\) in the order of the letters in \(U\). Put \(A=M_2,\ B=M_8,\ C=M_{13},\ D=M_0\). Replacing \(M_1\) by \(A=D^{-1}M_1D\) changes the root by a multiple of the root of \(D\) and the covector by a multiple of its covector, preserving both independent families. Let \(H\) and \(F\) denote the root and covector spans of \(A,B,C\), respectively. Both have dimension three, and in the pair \((v,\nu)\) of \(D\) one has \(v\notin H\) and \(\nu\notin F\). Rows \(1\) through \(7\) and row \(13\) propagate pairs in \(H,F\). Rows \(0\) and \(8\) give \(DAD^{-1}=X^{-1}DX\) for \(X=M_{10}\). If \(A,X\) have pairs \((a,f)\) and \((x,\xi)\) in \(H,F\), comparing the pairs on both sides gives \[ \begin{aligned} a-\nu(a)v &= \lambda\bigl(v+\mu^{-1}\xi(v)x\bigr),\\ f+\mu^{-1}f(v)\nu &= \lambda^{-1}\bigl(\nu-\nu(x)\xi\bigr) \end{aligned} \] for some \(\lambda\ne 0\). The coefficients on \(v,\nu\) give \(\lambda=-\nu(a)\) and \(f(v)=\mu\lambda^{-1}\). The other components give \(a=\lambda\mu^{-1}\xi(v)x\) and \(f=-\lambda^{-1}\nu(x)\xi\). Applying \(\nu\) to the expression for \(a\) gives \(\nu(x)\xi(v)=-\mu\). Thus \(af=x\xi\) and \(X=A\). The relation is now \(DAD=ADA\). Row \(9\), together with \(M_1=DAD^{-1}\), gives \(M_{11}=D\). Next, rows \(10\)--\(12\) send \((D,P)\) to \((P,N)\), then \((N,Q)\), then \((Q,C)\), where \(P=M_5,\ N=M_{12}\) and \(Q=M_6\). Write the pair of \(P\) as \((p,\pi)\) in \(H,F\), and put \(x_1=\pi(v),\ y_1=\nu(p),\ h_1=1+x_1y_1/\mu\). Conjugating at each step gives \[ \begin{array}{c|cc} &\text{root}&\text{covector}\\ \hline N&v+(x_1/\mu)p&\nu-y_1\pi\\ Q&y_1v+h_1p&-(x_1/\mu)\nu+h_1\pi\\ C&h_1v+(x_1/\mu)(1+h_1)p&h_1\nu-y_1(1+h_1)\pi. \end{array} \] For example \(N=P^{-1}DP\); the next row uses \(Q=N^{-1}PN\). The last pair lies in \(H,F\), so \(h_1=0\) and \(x_1y_1=-\mu\). Its remaining coefficients multiply to one, which gives \(P=C\). Of the eight rows supported in \(H,F\), rows \(1\)--\(7\) suffice for the contradiction. They now satisfy the hypotheses of Lemma~\ref{lem:seven} and contain the three independent pairs of \(A,B,C\). This proves the span bound. \end{proof} \begin{lemma}\label{lem:seven} Suppose the rank-one meridian matrices of common nonzero determinant \(\mu\) satisfy the seven signed relations in rows \(1\)--\(7\) of \eqref{eq:prefix}, with \(M_{10}=M_2\) and \(M_5=M_{13}\). Their root and covector spans cannot both have dimension three. \end{lemma} \begin{proof} Reorder the crossings so that the propagation comes first: \[ \begin{gathered} (9,14,15,+),\ (15,9,2,+),\ (8,2,3,-),\ (14,3,4,+),\ (3,14,13,-),\\ (4,9,8,+),\qquad (2,13,4,+). \end{gathered} \] Here the rows with over-arcs \(3,4,2\) and output arcs \(13,8,4\), respectively, solve the corresponding prefix relations for the other under-arc. The first five crossings propagate all the matrices from \(M_8,M_{14},M_9\). If both spans have dimension three, the corresponding pairs \((u,f),(v,g_0),(w,j)\) are separately independent. Set \[ \begin{gathered} x=f(v),\quad X=g_0(u),\quad y=f(w),\quad Y=j(u),\quad z=g_0(w),\quad Z=j(v),\qquad \eta=1-\mu,\\ h=1+zZ/\mu,\qquad t=-zx/\mu+hy,\qquad \ell=ZX+hY,\\ \kappa=tX/\mu+z(h-\eta/\mu),\qquad \Lambda=-\ell x+(\eta+h)Z . \end{gathered} \] In the bases \((u,v,w)\) and \((f,g_0,j)\), propagation gives \[ \begin{array}{c|cc} i&u_i&f_i\\\hline 2&(0,-z/\mu,h)&(0,Z,h)\\ 3&(t/\mu,-z/\mu,h)&(-\ell,Z,h)\\ 4 &(t/\mu,-z/\mu-\kappa,h)&(-\ell,Z+\Lambda/\mu,h)\\ 13&(\Lambda t/\mu^2,1-\Lambda z/\mu^2,\Lambda h/\mu) &(\kappa\ell,1-\kappa Z,-\kappa h). \end{array} \] For the first row, \(M_{15}=M_{9}M_{14}M_{9}^{-1}\) has pair \((v-Zw,g_0+(z/\mu)j)\), and \(M_{2}=M_{15}M_{9}M_{15}^{-1}\). The next two rows conjugate by \(M_{8}^{-1}\) and \(M_{14}\), respectively, with \(t=f(u_{2}),\ \ell=f_{2}(u)\) and \(\kappa=g_0(u_{3}),\ \Lambda=f_{3}(v)\). The last row uses \(M_{13}=M_{3}^{-1}M_{14}M_{3}\). At the crossing \((4,9,8,+)\), put \(P=f_{4}(w)\) and \(Q=j(u_{4})\). The first and third coefficients of \((M_{4}w,jM_{4}^{-1})=(\gamma u,\gamma^{-1}f)\) give \[ -Pt/\mu=\gamma,\qquad Ph=1,\qquad -Q\ell/\mu=\gamma^{-1},\qquad Qh/\mu=-1. \] Thus \(P,Q,t,\ell,h\) are nonzero. The middle coefficients give \(\kappa=-z/\mu\) and \(\Lambda=-\mu Z\). At \((2,13,4,+)\) the \(M_{2}\) pair has zero first coefficients. Comparing the first coefficients of the pairs of \(M_{13}\) and \(M_{2}^{-1}M_{4}M_{2}\) therefore gives \[ \Lambda t/\mu^2=\beta t/\mu,\qquad \kappa\ell=-\beta^{-1}\ell \] with \(\beta\ne0\). Cancelling \(t,\ell\) yields \(zZ=\kappa\Lambda=-\mu\). This makes \(h=0\), contrary to \(Ph=1\). \end{proof} For an irreducible representation in Cornwell's correspondence, the root span is nonzero and invariant, hence equals \(V\). The common kernel of the covectors is the fixed subspace and is zero, so the covectors span \(V^*\). Lemma~\ref{lem:spans} bounds the dimension by \(b-1\) and proves the augmentation inequality in Theorem~\ref{thm:main}.