\chapter{Uniqueness of the infinite-volume frustration-free state}\label{s:uniqueness} The coupling of \zcref{s:block-coupling} also controls the dependence of the spins in a fixed set on the boundary configuration of a growing box. We use this fact and the polynomial description of the ground space to prove uniqueness on the infinite square lattice. For a finite set $A \subset \Z^2$, let $\mathcal H_A = \bigotimes_{x \in A} V_x$, with the spin-$2$ spaces of \zcref{s:quantum-poly}. We regard an operator on $\mathcal H_A$ as an operator on $\mathcal H_{A'}$ for $A \subset A'$ by tensoring it with the identity. An infinite-volume state is a state on the norm closure of the union of these local matrix algebras. We call the state $\omega$ \emph{frustration-free} if $\omega(P_e) = 0$ for every bond $e$ of $\Z^2$. \begin{theorem}\label{t:ff-unique} The spin-$2$ AKLT model on $\Z^2$ has a unique infinite-volume frustration-free state. \end{theorem} \section{The central marginal of the classical law}\label{s:uniqueness-mixing} Let $A \subset \Z^2$ be finite and nonempty, and set $R = \max_{x \in A} |x|_\infty$. Recall the box $\mathcal C_M$, the length $\ell = M^{1/4}$, and the number of trials $J = \lfloor M^{1/8} \rfloor$. By \zcref{t:inputs,t:contraction,t:window}, Hypothesis~(W) holds, so we may use the coupling of \zcref{s:block-coupling} with $\delta = p_0 \varepsilon_{192} > 0$. For all sufficiently large $M$, we have $A \subset \mathcal C_M$ and $M + 1 - R > (J + 5) \ell$, since $(J + 5) \ell / M \to 0$. Suppose that two boundary configurations of $\mathcal C_M$ differ at just one site $y$ adjacent to $\mathcal C_M$. In the coupling, the probability that none of the trials succeeds is at most $(1 - \delta)^J$, by \eqref{e:nosuccess}. If some trial at $i$ succeeds, then the final configurations agree on $F_{i + 1}$, by the agreement argument in the proof of \zcref{p:coupling}. Since $|y|_\infty = M + 1$, for all $x \in A$ we have \begin{equation*} |x - y|_\infty \ge M + 1 - R > (J + 5) \ell \ge i \ell , \end{equation*} where we used \eqref{e:indexbound}. Thus $A \subset F_{i + 1}$, and the two marginals on $A$ differ in total variation by at most $(1 - \delta)^J$. The law $\mu^\eta_{\mathcal C_M}$ depends on $\eta$ only at the $4(2M + 1)$ exterior sites adjacent to $\mathcal C_M$. For two arbitrary boundary configurations $\eta$ and $\eta'$, change these spins from $\eta$ to $\eta'$ one at a time. By the triangle inequality for total variation, \begin{equation}\label{e:unique-tv} \sup_{\eta, \eta'} \bigl\| \mu^\eta_{\mathcal C_M,A} - \mu^{\eta'}_{\mathcal C_M,A} \bigr\|_{\mathrm{TV}} \le 4(2M + 1)(1 - \delta)^J \longrightarrow 0 \qquad \text{as } M \to \infty . \end{equation} Here the total variation distance of two probability measures is the supremum, over measurable events, of the absolute difference of their probabilities. For each such event, \eqref{e:unique-tv} bounds the variation of its probability over all boundary configurations. The same bound thus holds between any two mixtures of these marginals. \section{Coherent-state laws of open-box ground states}\label{s:uniqueness-coherent} For a finite set $A$, let $L_s$ be the evaluation functional $\varphi \mapsto \varphi(\zeta(s))$ on $\mathcal H_A$, for $s \in (\Sph)^A$, and let $E_A(s) = L_s^* L_s$. By \eqref{e:Vinner} and the tensor product construction of the inner product, \begin{equation*} \int E_A(s) \prod_{x \in A} 5 \dd s_x = \Id_{\mathcal H_A} . \end{equation*} For a density matrix $\rho$ on $\mathcal H_A$, the \emph{coherent-state law} of $\rho$ is the probability law with density $\Tr(\rho E_A(s))$ with respect to $\prod_{x \in A} 5 \dd s_x$. In particular, for a unit vector $\varphi$, the coherent-state law of its rank-one projection has density $|\varphi(\zeta(s))|^2$. For $A \subset A'$, the operator $E_{A'}(s)$ is the tensor product of $E_A(s_A)$ and $E_{A'\setminus A}(s_{A'\setminus A})$. Given a density matrix on $\mathcal H_{A'}$, integrating out $s_{A'\setminus A}$ in its coherent-state law therefore gives the coherent-state law of its restriction to $A$. The open box we use is $\mathcal C_{M + 1}$. Let $\mathcal F_M$ be the set of all bonds with both sites in $\mathcal C_{M + 1}$, and set $H^{\mathrm{open}}_M = \sum_{e \in \mathcal F_M} P_e$. If $\varphi$ is a unit vector in its kernel, then $P_e \varphi = 0$ for every $e \in \mathcal F_M$ because the terms are nonnegative. The divisibility argument of \zcref{l:edge,l:divisibility} applies to these bonds. Indeed, the proof of \zcref{l:edge} only uses the tensor factors at the ends of a bond, and the polynomials $q_e$ for distinct bonds are pairwise non-associated irreducibles, as in the proof of \zcref{l:divisibility}. Thus \begin{equation*} \varphi = q_{\mathcal F_M} \chi , \end{equation*} where $\chi$ is a polynomial in the variables at sites of $\mathcal C_{M + 1} \setminus \mathcal C_M$. To see the last assertion, every site of $\mathcal C_M$ belongs to four bonds of $\mathcal F_M$, so the product $q_{\mathcal F_M}$ uses all four degrees there. By \eqref{e:qmod}, the coherent-state density of this vector is proportional to \begin{equation*} |\chi(\zeta(s))|^2 \prod_{\{x, y\} \in \mathcal F_M} (1 - s_x \cdot s_y) . \end{equation*} Make the change of variables $n_x = (-1)^{x_1 + x_2} s_x$ on $\mathcal C_{M + 1}$. Each bond factor then becomes $1 + n_x \cdot n_y$. Conditional on the spins of the outer layer $\mathcal C_{M + 1} \setminus \mathcal C_M$, the factor from $\chi$ and all bond factors within that layer are constant. Every bond of $\Z^2$ that meets $\mathcal C_M$ belongs to $\mathcal F_M$. A version of the conditional law of $n_{\mathcal C_M}$ is therefore $\mu^\eta_{\mathcal C_M}$ of \eqref{e:muA}, where $\eta$ agrees with the transformed outer-layer spins on the sites adjacent to $\mathcal C_M$ and is arbitrary on the other exterior sites. Its density is obtained by normalizing the remaining bond factors, whose integral is positive for every $\eta$ by \zcref{s:quantum-rotated}. It follows that the law of $n_A$ for $A \subset \mathcal C_M$ is a mixture of $\mu^\eta_{\mathcal C_M,A}$. The same is true of any density matrix supported in $\ker H^{\mathrm{open}}_M$, by its decomposition into a convex combination of rank-one projections in the kernel. Thus \eqref{e:unique-tv} applies to any two such density matrices. Taking the inverse change of signs on $A$ preserves total variation distance, so the bound applies also to the coherent-state laws of their restrictions to $A$. \section{Determination of the state}\label{s:uniqueness-state} For each finite $A$, the coherent-state law determines the density matrix on $\mathcal H_A$. To see this, let $B$ be the difference of two density matrices with the same coherent-state law. Then $\Tr(B E_A(s)) = 0$ almost everywhere. Choose the orthonormal basis of $V$ formed by $\binom{4}{k}^{1/2} z_1^{4 - k} z_2^k$ for $0 \le k \le 4$, using \eqref{e:explicit-gram}, and its tensor product on $A$. On the chart in which the first spinor component is nonzero at every site, use the representatives $z_x = (1, t_x)/\sqrt{1 + |t_x|^2}$ with $t_x \in \C$. The choice of phases of these representatives does not change $E_A(s)$, since all polynomials have the same degree in each $z_x$. On this chart, $\Tr(B E_A(s))$ is continuous, and the measure $\prod_{x \in A} \dd s_x$ has full support. Hence it vanishes throughout the chart. Multiplying by $\prod_{x \in A} (1 + |t_x|^2)^4$ yields a polynomial in the monomials \begin{equation*} \prod_{x \in A} t_x^{k_x} \overline{t_x}^{l_x} , \qquad 0 \le k_x, l_x \le 4 . \end{equation*} Their coefficients are the entries of $B$ in the tensor basis, each multiplied by a nonzero factor. These monomials are linearly independent as functions of $(t_x)_{x \in A}$, since the formal variables $t_x$ and $\overline{t_x}$ are obtained by an invertible complex linear change of variables from $\Re t_x$ and $\Im t_x$. Thus $B = 0$. \begin{proof}[Proof of \zcref{t:ff-unique}] Let $\omega$ be an infinite-volume frustration-free state. Its restriction to $\mathcal C_{M + 1}$ has a density matrix $\rho_M$ with $\Tr(\rho_M H^{\mathrm{open}}_M) = 0$. Since both operators in this trace are nonnegative, $\rho_M$ is supported in $\ker H^{\mathrm{open}}_M$. Indeed, each eigenvector of $\rho_M$ with positive eigenvalue has zero expectation of $H^{\mathrm{open}}_M$ and lies in its kernel. If $\omega'$ is another such state, apply \eqref{e:unique-tv} and the argument of \zcref{s:uniqueness-coherent} to the restrictions of $\omega$ and $\omega'$ to $\mathcal C_{M + 1}$. For every fixed finite $A$, their coherent-state laws on $A$ are independent of $M$ as soon as $A \subset \mathcal C_M$. The bound in \eqref{e:unique-tv} tends to zero, so these two laws are equal. The density matrices on $A$ are equal by the preceding argument. The local matrix algebras have dense union, and we conclude that $\omega = \omega'$. For existence, the open-box kernel is nonzero for every $M$. For $x \in \mathcal C_{M + 1}$, let $d_x$ be the number of bonds of $\mathcal F_M$ at $x$. The nonzero polynomial $q_{\mathcal F_M} \prod_{x \in \mathcal C_{M + 1}} z_{x,1}^{4 - d_x}$ belongs to $\mathcal H_{\mathcal C_{M + 1}}$ and lies in $\ker H^{\mathrm{open}}_M$ by \zcref{l:edge}. Normalize it and take its vector state. Along a diagonal subsequence of growing boxes, the restrictions of these states to each fixed finite box converge, by compactness of the finite-dimensional sets of density matrices. The limits are consistent under restriction, so they give a state on the local algebras and hence on their norm closure. For each bond $e$, the finite-volume expectations of $P_e$ are zero once both sites belong to the box. The limit state is therefore frustration-free. \end{proof}