\begin{wrapfigure}{R}{.46\textwidth} \centering \includegraphics[width=\linewidth]{assets/cf12f5dc1cb5-L0-cut-word.pdf} \caption{The red line marks the regular cut. The portion to its left is $L(\mathcal L_0)$, and the portion to its right is $R(\mathcal L_0)$.}\label{fig:trefoil-cut-word} \end{wrapfigure} View the box from the side and rotate the view about the $z$-axis through $\arctan(1/4)$, keeping the inserted $\gamma$ at $z=1/2$. Its disk contains the puncture points $(1,1)$ and $(5,1)$. The other two intersections lie near the faces $y=0,2$, outside the rectangle. The side-view construction of Theorem~\ref{thm:box-insertion-cut} gives the cut colors $(1,a,a,1)$, with the middle pair precisely the block enclosed by $\gamma$. We apply relative Reidemeister moves separately to the two halves, fixing every ordered endpoint at the cut and at the exterior boundary, and preserving the strand colors. Types II and III preserve the blackboard framing, while type I corrections are recorded using Figure~\ref{fig:framing-curl}. The half lengths decrease from $3,8$ to $3,3$. The left half has no framing change, and restoring the right half's framing contributes $-t^{-3}$. We include this scalar in the right accumulator. The moving component winds once around each puncture, with opposite signs. Each winding requires at least two mixed crossings, and the closed component requires a cup and a cap. Thus six letters are necessary, and the resulting words attain this bound: \[ L(\mathcal L_0)=\MorseColorBoxes{a_{1}^{2}}\MorseCup{2}{3}\MorseColorBoxes{a_{1}\\\hline a_{4}}\MorseUnder{1}{2}{1}{2}\MorseColorBoxes{a_{2}\\\hline a_{4}}\MorseUnder{3}{4}{3}{4}\MorseColorBoxes{a_{2}^{2}},\qquad R(\mathcal L_0)=\MorseColorBoxes{a_{2}^{2}}\MorseUnder{1}{2}{1}{2}\MorseColorBoxes{a_{1}\\\hline a_{3}}\MorseOver{2}{3}{2}{3}\MorseColorBoxes{a_{1}^{2}}\MorseCap{3}{4}\MorseColorBoxes{a_{1}^{2}}. \] Start with $\mathbf a_0=(a,a)$, so both punctures have color $a$ and $\alpha'=\alpha$, as in Section~\ref{sec:algebraic-insertions}. Admissibility of $(a,a,c)$ requires $0\le c\le2a$ and $2a+c\in2\mathbb Z$, so $c$ is even. Write $c=2b$ and put $u=t^a$ and $v=t^b$. We work in the interior admissible range $2\le c\le2a-2$. The input set is $\mathbf L_0=\{(c)\}$ and $v_0=(1)$. The first letter is $\MorseCup{2}{3}$, giving $\mathbf a_1=(a,1,1,a)$. Label the successive internal edges of its left-associated fusion tree by $s_1,s_2,c$. Thus \[ \mathbf L_1=\left\{(s_1,s_2,c)\in\mathbb N^3\ \middle|\ \begin{array}{l} (a,1,s_1),\ (s_1,1,s_2),\ (s_2,a,c)\\ \text{and }(a,a,c)\text{ are admissible} \end{array}\right\}. \] We take the colors to be symbolic and sufficiently large within the admissible region, with $c$ and $2a-c$ sufficiently large as well. For this step it suffices that $a\geq2$ and $2\leq c\leq2a-2$. The joint admissibility system is \[ \begin{aligned} |a-1|&\leq s_1\leq a+1, & a+1+s_1&\equiv0\pmod2,\\ |s_1-1|&\leq s_2\leq s_1+1, & s_1+1+s_2&\equiv0\pmod2,\\ |a-c|&\leq s_2\leq a+c, & a+c+s_2&\equiv0\pmod2,\\ 0&\leq c\leq2a, & 2a+c&\equiv0\pmod2. \end{aligned} \] The first interval contains $a-1,a,a+1$, and parity excludes $a$. Thus $s_1=a-1$ or $s_1=a+1$. The second interval contains $s_1-1,s_1,s_1+1$, and parity excludes $s_1$. The four resulting pairs satisfy the third and fourth rows under our assumptions. Therefore \[ \mathbf L_1=\{(a-1,a-2,c),\ (a-1,a,c),\ (a+1,a,c),\ (a+1,a+2,c)\}. \] There is one input labeling $m=(c)$, so we compute the four entries $V_1(\ell,(c))$, in the order just displayed for $\mathbf L_1$. The inserted pair has joining color $0$. Fusing this pair with the first strand of color $a$ gives the internal edge whose possible labels in $\mathbf L_1$ are $a-2,a,a+2$. The triple $(a,0,a)$ is admissible. The row $(a-1,a-2,c)$ would require $(a,0,a-2)$, which fails $a\leq a-2$. The row $(a+1,a+2,c)$ would require $(a,0,a+2)$, which fails $a+2\leq a$. The fusion rules therefore assign zero to both entries, giving \[ V_1((a-1,a-2,c),(c))=0,\qquad V_1((a+1,a+2,c),(c))=0. \] For the two remaining rows, the cup gives the tree $(a,(1,1)_0)_a$, whereas our coordinate tree is $((a,1)_{a\pm1},1)_a$. The final strand of color $a$ and the root $c$ are unchanged. Reading this local picture from right to left, \eqref{eq:recoupling} has boundary colors $(1,1,a)$, total color $a$, initial channel $0$, and new channel $a-1$ or $a+1$. Thus \[ V_1((a\pm1,a,c),(c))= \begin{Bmatrix}1&1&a\pm1\\a&a&0\end{Bmatrix} =\frac{\Delta_{a\pm1} \left\langle\begin{matrix}1&1&0\\a&a&a\pm1\end{matrix}\right\rangle} {\theta(1,a,a\pm1)^2}. \] For the channel $a-1$, the four lower bounds in the tetrahedron sum \eqref{eq:tetrahedron-factorials} are $(1,a,a,a)$ and the three upper bounds are $(a+1,a,a)$. Only the summation index $a$ remains. For the channel $a+1$, the bounds are $(1,a+1,a,a+1)$ and $(a+1,a+1,a+1)$, so only the index $a+1$ remains. Cancelling the quantum factorials gives \[ \begin{aligned} \left\langle\begin{matrix}1&1&0\\a&a&a-1\end{matrix}\right\rangle &=\frac1{[a]}\,(-1)^a[a][a+1]=(-1)^a[a+1],\\ \left\langle\begin{matrix}1&1&0\\a&a&a+1\end{matrix}\right\rangle &=\frac1{[a+1]}\,(-1)^{a+1}[a+1][a+2]=(-1)^{a+1}[a+2]. \end{aligned} \] The theta formula similarly gives $\theta(1,a,a-1)=(-1)^a[a+1]$ and $\theta(1,a,a+1)=(-1)^{a+1}[a+2]$. Using $\Delta_{a-1}=(-1)^{a-1}[a]$ and $\Delta_{a+1}=(-1)^{a+1}[a+2]$, the two entries are therefore \[ \begin{aligned} V_1((a-1,a,c),(c)) &=\frac{(-1)^{a-1}[a](-1)^a[a+1]}{[a+1]^2} =-\frac{[a]}{[a+1]}=-\frac{t^2(u^4-1)}{t^4u^4-1},\\ V_1((a+1,a,c),(c)) &=\frac{(-1)^{a+1}[a+2](-1)^{a+1}[a+2]}{[a+2]^2}=1. \end{aligned} \] In the displayed order of $\mathbf L_1$, $v_1^T=(V_1v_0)^T=\bigl(0,-\frac{t^2(u^4-1)}{t^4u^4-1},1,0\bigr)$. The second letter is $\MorseUnder{1}{2}{1}{2}$, giving $\mathbf a_2=(1,a,1,a)$. Exchanging the first two boundary colors leaves the admissibility conditions unchanged. Hence $\mathbf L_2=\mathbf L_1$, and the crossed strands are already fused together in the chosen tree, so no change of tree is needed. The undercrossing uses the reciprocals of the crossing coefficients in \eqref{eq:colored-crossing}, therefore we have $(R^{a,1}_{a-1})^{-1}=-t^2u$ and $(R^{a,1}_{a+1})^{-1}=u^{-1}$. With rows in the displayed order of $\mathbf L_2$ and columns in the same order of $\mathbf L_1$, $V_2=\operatorname{diag}(-t^2u,-t^2u,u^{-1},u^{-1})$. The third letter is $\MorseUnder{3}{4}{3}{4}$, giving $\mathbf a_3=(1,a,a,1)$. Its admissible labelings satisfy $(1,a,s_1)$, $(s_1,a,s_2)$, $(s_2,1,c)$ and $(a,a,c)$. The first and third conditions give $s_1=a\pm1$ and $s_2=c\pm1$, and all four combinations satisfy the remaining conditions in our range. Thus \[ \mathbf L_3=\{(a-1,c-1,c),(a-1,c+1,c),(a+1,c-1,c),(a+1,c+1,c)\}. \] For this crossing, recouple the last pair, apply the inverse crossing coefficient, and recouple back, as in Theorem~\ref{thm:fusion-evaluation}. With columns in the order of $\mathbf L_2$ and rows in the order of $\mathbf L_3$, the evaluated transition matrix is \[V_3=\begin{pmatrix}\frac{t^{4} v^{4} - 1}{u \left(t^{4} v^{8} - 1\right)}&\frac{u \left(v^{4} - 1\right) \left(t^{4} u^{4} v^{4} - 1\right)}{v^{2} \left(u^{4} - 1\right) \left(t^{4} v^{8} - 1\right)}&0&0\\\frac{t^{2} v^{2}}{u}&\frac{t^{2} u \left(- u^{4} + v^{4}\right)}{u^{4} - 1}&0&0\\0&0&\frac{v^{4} - 1}{u \left(t^{4} v^{8} - 1\right)}&\frac{t^{2} u \left(t^{4} v^{4} - 1\right) \left(t^{8} u^{4} v^{4} - 1\right)}{v^{2} \left(t^{4} v^{8} - 1\right) \left(t^{8} u^{4} - 1\right)}\\0&0&\frac{v^{2}}{u}&\frac{t^{6} u \left(- t^{4} u^{4} + v^{4}\right)}{t^{8} u^{4} - 1}\end{pmatrix}.\] Applying $V_3V_2v_1$ gives the left accumulator \[v_3(L(\mathcal L_0))^T=\left(\frac{t^{4} u^{2} \left(v^{4} - 1\right) \left(t^{4} u^{4} v^{4} - 1\right)}{v^{2} \left(t^{4} u^{4} - 1\right) \left(t^{4} v^{8} - 1\right)},\frac{t^{6} u^{2} \left(- u^{4} + v^{4}\right)}{t^{4} u^{4} - 1},\frac{v^{4} - 1}{u^{2} \left(t^{4} v^{8} - 1\right)},\frac{v^{2}}{u^{2}}\right).\] For $R(\mathcal L_0)^*$, start again at $(a,a)$ with $v_0=(1)$. Reversing the right word interchanges its cap with a cup and gives the three updates $\MorseCup{3}{4}$, $\MorseOver{2}{3}{2}{3}$ and $\MorseUnder{1}{2}{1}{2}$. Their color columns are $(a,a,1,1)$, $(a,1,a,1)$ and $(1,a,a,1)$. Applying the same fusion algorithm and the framing correction $-t^{-3}$ gives, in the displayed order of $\mathbf L_3$,\footnote{The left and right accumulators coincide up to a framing twist in this example. This is a coincidence of this knot and this cut. At a different cut, or for a different knot, the two accumulators need not agree up to a twist.} $v_3(R(\mathcal L_0)^*)^T=-t^{-3}v_3(L(\mathcal L_0))^T$. The insertion block is the middle pair $(a,a)$. Write $L=v_3(L(\mathcal L_0))$ and $R=v_3(R(\mathcal L_0)^*)$. As in Theorem~\ref{thm:gamma-insertion}, choose the tree that first fuses this pair. Its admissible label set is \[ \mathbf L=\{(s_1,s_2,c)\in\mathbb N^3\mid (a,a,s_1),\ (1,s_1,s_2),\ (s_2,1,c)\text{ and }(a,a,c)\text{ are admissible}\}. \] The condition $(s_2,1,c)$ gives $s_2=c\pm1$, and $(1,s_1,s_2)$ gives $s_1=s_2\pm1$. The condition $(a,a,s_1)$ retains all four resulting labelings in our range, so \[ \mathbf L=\{(c-2,c-1,c),(c,c-1,c),(c,c+1,c),(c+2,c+1,c)\}. \] The output edge of the subtree with leaves $(a,a)$ has label $s_1$, so $e(\ell)=s_1$. Thus the map of Theorem~\ref{thm:gamma-insertion} is explicitly \[ \begin{aligned} e((c-2,c-1,c))&=c-2,& e((c,c-1,c))&=c,\\ e((c,c+1,c))&=c,& e((c+2,c+1,c))&=c+2. \end{aligned} \] With columns in the displayed order of $\mathbf L_3$ and rows in that of $\mathbf L$, the change-of-basis matrix $C=C_c$ of Theorem~\ref{thm:cut-recombination}, obtained from \eqref{eq:recoupling}, is \[ C=\begin{pmatrix} -\frac{v^2}{v^4+1}&0&\frac{t^4u^4v^4-1}{(t^4u^4-1)(v^4+1)}&0\\ 1&0&\frac{t^4u^4-v^4}{v^2(t^4u^4-1)}&0\\ 0&-\frac{t^2v^2}{t^4v^4+1}&0&\frac{t^8u^4v^4-1}{(t^4u^4-1)(t^4v^4+1)}\\ 0&1&0&\frac{t^2(u^4-v^4)}{v^2(t^4u^4-1)} \end{pmatrix}. \] It changes $((1,a)_{a\pm1},a)_{s_2}$ into $(1,(a,a)_{s_1})_{s_2}$. The bubble identity gives the original basis norms, for\newline \mbox{$\ell=(s_1,s_2,c)\in\mathbf L_3$}, as $n_{c,\ell}=\theta(1,a,s_1)\theta(s_1,a,s_2)\theta(s_2,1,c)/ (\Delta_{s_1}\Delta_{s_2}\Delta_c)$. By Theorems~\ref{thm:cut-recombination} and~\ref{thm:gamma-insertion}, \[ \langle\mathcal L_j\rangle(\chi_a,\chi_c) =\frac1{\nu_{a,c}}\sum_{\ell\in\mathbf L_3}n_{c,\ell}L_\ell \left(C^{-1}\operatorname{diag}(\chi_{c-2}^j,\chi_c^j,\chi_c^j,\chi_{c+2}^j)CR\right)_\ell. \] For $j=0$, this gives $\langle\mathcal L_0\rangle(\chi_a,\chi_c) =t\chi_c+t^{-1}\chi_a^2$.