\chapter{Explicit constants}\label{s:explicit} In this appendix, we make the constants in the proof of \zcref{t:main} explicit, and we derive the values $L_0 = 2 \lceil \exp(10^{40}) \rceil + 5$ and $c = \exp[-\exp(\exp(10^{41}))]$ of \eqref{e:mainconstants}. We follow the steps of the proof in \zcref{s:computer-proofmain}. The constant $c$ is dominated by the dimension of the quantum patch of side $b$, so that coarse estimates suffice in the earlier steps. In \zcref{s:explicit-table}, we list the constants of the paper and where they come from. The numerical inequalities below involve only rational numbers and elementary functions, and each of them can be checked by hand. \section{The constants of Hypothesis (P)}\label{s:explicit-P} We first make the constants of \zcref{l:Ktheta,t:contraction} explicit. \begin{lemma}\label{l:explicitP} Hypothesis~(P) holds with \begin{equation}\label{e:explicitA} A = 10^{13}, \qquad \theta = 1 - 10^{-20} . \end{equation} \end{lemma} \begin{proof} We follow the proofs of \zcref{l:Ktheta,t:contraction}, keeping track of the constants. \emph{A bound for the masses.} \zcref[S]{t:inputs} shows that Hypotheses (H1), (H2), and (H3) hold at $(\cs, \ct) = (961/1000, 187/200)$, with $\gamma_{\mathrm o} = 2499/2500$, $C_{\mathrm o} = 10$, and $K_{\mathfrak t}(1) \le 0.92320$ for both types. Recall the bound \eqref{e:massmajorant} from the proof of \zcref{l:Ktheta}, which states that for all $n \ge 1$, \begin{equation}\label{e:explicit-mass} \frac{\mathsf m_{\mathfrak t}(n)}{h_{\mathfrak t}} \le 2 n C_{\mathrm o} \gamma_{\mathrm o}^n . \end{equation} \emph{Continuity in $t$.} Let $\theta \in [0.9999, 1)$. The mean value theorem shows that $\theta^{-n - 1} - 1 \le (n + 1)(1 - \theta) \theta^{-n - 2}$ for $n \ge 1$. Inserting this and \eqref{e:explicit-mass} into \eqref{e:kernel}, we obtain \begin{equation}\label{e:explicit-Kdiff} K_{\mathfrak t}(\theta) - K_{\mathfrak t}(1) = \frac{1}{h_{\mathfrak t}} \sum_{n \ge 1} \mathsf m_{\mathfrak t}(n) (\theta^{-n - 1} - 1) \le 2 C_{\mathrm o} (1 - \theta) \theta^{-2} \sum_{n \ge 1} n (n + 1) \Bigl( \frac{\gamma_{\mathrm o}}{\theta} \Bigr)^n . \end{equation} Recall that $\sum_{n \ge 1} n (n + 1) r^n = 2r/(1 - r)^3$ for $0 < r < 1$. Since $\gamma_{\mathrm o}/\theta \le 0.9996/0.9999 = 3332/3333$, we have \begin{equation*} \sum_{n \ge 1} n (n + 1) \Bigl( \frac{\gamma_{\mathrm o}}{\theta} \Bigr)^n \le 2 \cdot 3333^3 < 10^{11} . \end{equation*} Using also $\theta^{-2} < 2$ and $C_{\mathrm o} = 10$, we deduce from \eqref{e:explicit-Kdiff} that $K_{\mathfrak t}(\theta) - K_{\mathfrak t}(1) \le 4 \cdot 10^{12} (1 - \theta)$. For $\theta = 1 - 10^{-20}$, this gives \begin{equation}\label{e:explicit-Ktheta} K_\theta = \max_{\mathfrak t} K_{\mathfrak t}(\theta) \le 0.92320 + 4 \cdot 10^{-8} < 0.9233 . \end{equation} \emph{The constant $A$.} Recall that $C_{\mathrm{pin}} = 2 C_{\mathrm o}/(1 - \gamma_{\mathrm o})$. The proof of \zcref{t:contraction} uses only that $\theta \in (\gamma_{\mathrm o}, 1)$ and $K_\theta < 1$. For every such $\theta$, it gives Hypothesis~(P) with this $\theta$ and \begin{equation}\label{e:explicit-Aformula} A = \frac{h_{\max}}{h_{\min}} \max \Bigl( 2, \frac{C_{\mathrm{pin}}}{1 - K_\theta} \Bigr) . \end{equation} Now let $\theta = 1 - 10^{-20}$, which is larger than $\gamma_{\mathrm o}$. We have $C_{\mathrm{pin}} = 5 \cdot 10^4$, and $1 - K_\theta > 0.0767$ by \eqref{e:explicit-Ktheta}. Therefore the right side of \eqref{e:explicit-Aformula} satisfies \begin{equation*} \frac{h_{\max}}{h_{\min}} \max \Bigl( 2, \frac{C_{\mathrm{pin}}}{1 - K_\theta} \Bigr) \le \frac54 \cdot \frac{5 \cdot 10^4}{0.0767} < 10^6 . \end{equation*} Hypothesis~(P) with a constant $A$ implies (P) with every larger constant, so that (P) holds with the values \eqref{e:explicitA}. \end{proof} \section{The window and block estimates}\label{s:explicit-block} Throughout the rest of this appendix, we set \begin{equation}\label{e:explicitM} M = \lceil \exp(10^{40}) \rceil, \qquad b = 2M + 1, \end{equation} and $\ell = M^{1/4}$, as in \zcref{s:block}. We now check that this $M$ satisfies the requirements of \zcref{t:window} and \zcref{p:blockgap}, with $A$ and $\theta$ as in \eqref{e:explicitA}. Recall the threshold $M_{\mathrm g} = 73^4$ of \zcref{l:shellgeom}, which is also used in \zcref{s:cut-geometry}. \begin{lemma}\label{l:explicitW} Let $A$ and $\theta$ be as in \eqref{e:explicitA}. Then Hypothesis~(W) holds with \begin{equation}\label{e:explicitW} M_\star = 10^{180}, \qquad p_0 = \exp(-2 \cdot 10^{36}) . \end{equation} \end{lemma} \begin{proof} The proof of \zcref{t:window} shows that (W) holds with $p_0 = e^{-2 C_{\mathrm{cut}}}$, where $C_{\mathrm{cut}}$ is given by \eqref{e:Ccut}, and with $M_\star = \max(M_{\mathrm g}, M_1)$, for every $M_1$ such that \eqref{e:Mone} holds for all $M \ge M_1$. We first show that we may take $M_1 = 10^{180}$. We claim that \begin{equation}\label{e:explicit-SM} (2M + 1)^2 \theta^{\ell/12 - 7} \le 1 \qquad \text{for all } M \ge (256 \cdot 10^{20})^8 . \end{equation} Using $2M + 1 \le 3 \ell^4$, $\theta^{-7} < 8$, and $\theta^{\ell/12} \le \exp(-10^{-20} \ell/12)$, we obtain \begin{equation*} (2M + 1)^2 \theta^{\ell/12 - 7} \le 72 \, \ell^8 \exp(-10^{-20} \ell/12) . \end{equation*} For $\ell \ge (256 \cdot 10^{20})^2$, we have $\sqrt \ell \ge 256 \cdot 10^{20}$, and therefore \begin{equation*} \log 72 + 8 \log \ell \le 9 \sqrt \ell < \frac{256}{12} \sqrt \ell \le \frac{10^{-20} \ell}{12} , \end{equation*} which proves \eqref{e:explicit-SM}. Since $(256 \cdot 10^{20})^8 = 2^{64} \cdot 10^{160} < 1.9 \cdot 10^{179}$, it follows that \eqref{e:Mone} holds for all $M \ge 10^{180}$. The threshold $M_{\mathrm g} = 73^4$ is smaller than $10^{180}$, and we may therefore take $M_\star = 10^{180}$. Next, inserting $A = 10^{13}$, $1/(1 - \theta) = 10^{20}$, and $C_\star = \log 8192 < 10$ into \eqref{e:Ccut}, we obtain \begin{equation}\label{e:explicit-Ccut} C_{\mathrm{cut}} < 10 + 480 \cdot 10^{13} (7 + 10^{20}) + 10^{14} < 10^{36} . \end{equation} This gives $p_0 = e^{-2 C_{\mathrm{cut}}} > \exp(-2 \cdot 10^{36})$. Hypothesis~(W) with a constant $p_0$ implies (W) with every smaller positive constant, which gives \eqref{e:explicitW}. \end{proof} \paragraph{The coupling.} By the definition \eqref{e:overlap} of $\varepsilon_K$, we have $\varepsilon_{192} = (3072)^{-768}/2 \ge 2^{-8899}$, since $3072 < 2^{11.585}$. With $p_0$ as in \eqref{e:explicitW}, the success probability of one coupling trial satisfies \begin{equation}\label{e:explicitdelta} \delta = p_0 \varepsilon_{192} \ge \exp(-2 \cdot 10^{36}) \, 2^{-8899} \ge \exp(-5 \cdot 10^{36}) . \end{equation} For $M$ as in \eqref{e:explicitM}, we have $M^{1/8} \ge \exp(10^{40}/8)$ and $J = \lfloor M^{1/8} \rfloor$, so that $M^{1/8}/2 \le J$ and $J + 5 \le 2 M^{1/8}$. The first term of $\Delta_M$ in \eqref{e:discrepancy} therefore satisfies \begin{equation*} 9 M^{1/2} (J + 5)^2 \le 36 M^{3/4} \le \frac M8 , \end{equation*} where we used that $M^{1/4} \ge 288$. For the second term, we use \eqref{e:explicitdelta} and $\log M \le 10^{40} + 1$ to find that \begin{equation*} \delta J \ge \frac12 \exp \Bigl( \frac{10^{40}}{8} - 5 \cdot 10^{36} \Bigr) \ge \frac12 \exp(10^{39}) \ge 4 \log M . \end{equation*} Using $b \le 3M$, we deduce that \begin{equation*} b^2 (1 - \delta)^J \le b^2 e^{-\delta J} \le b^2 M^{-4} \le 9 M^{-2} \le \frac M8 . \end{equation*} Combining the two bounds, we conclude that \begin{equation}\label{e:explicit-Delta} \Delta_M \le \frac M4 < \frac b8 . \end{equation} \paragraph{The block dynamics.} The value \eqref{e:explicitM} satisfies $M \ge 10^{180} = \max(M_\star, M_{\mathrm g})$, since $\log M \ge 10^{40} > 180 \log 10$. Therefore \zcref{p:blockgap} applies, and its bound \eqref{e:blockgap} and \eqref{e:explicit-Delta} imply that \begin{equation}\label{e:explicit-blockgap} \gap(\Lbl) \ge 1 - \frac{4 \Delta_M}{b} \ge \frac12 \qquad \text{for all } L, N \ge b + 4 . \end{equation} \section{The patch gap}\label{s:explicit-patch} Recall from \zcref{s:quantum} that the patch operator $h_b$ is the sum of the projections $P_e$ over the bonds $e$ of $\Z^2$ that meet $\{1, \dots, b\}^2$, on the space of the patch, and that $g_b$ is the smallest positive eigenvalue of $h_b$. We let \begin{equation}\label{e:explicit-patchsize} n_b = b^2 + 4b , \qquad m_b = 2b^2 + 2b , \qquad \dim_b = 5^{n_b} \end{equation} denote the number of sites of the patch, the number of bonds that meet $\{1, \dots, b\}^2$, and the dimension of the space on which $h_b$ acts, respectively. Let $\varphi_k = z_1^{4 - k} z_2^k$ for $0 \le k \le 4$, so that $(\varphi_k)_{0 \le k \le 4}$ is the monomial basis of $V$. \begin{lemma}\label{l:patchgapexplicit} For every integer $b \ge 1$, we have \begin{equation}\label{e:patchgapexplicit} g_b \ge [140 b (b + 1)]^{-5^{b^2 + 4b}} . \end{equation} \end{lemma} \begin{proof} We show that $70 h_b$ is an integer matrix in a suitable basis, and we then use its characteristic polynomial. \emph{The monomial basis.} By the definition \eqref{e:Vinner} of $\langle \cdot, \cdot \rangle_V$, we have \begin{equation}\label{e:explicit-gram} \langle \varphi_k, \varphi_l \rangle_V = \1_{k = l} \binom 4k^{-1} , \qquad 0 \le k, l \le 4 . \end{equation} Indeed, the monomials are orthogonal, since a rotation about the third axis multiplies $\varphi_k$ evaluated at a coherent state by a phase that depends on $k$. Further, if $a$ is uniformly distributed on $\Sph$, then $|\zeta_1(a)|^2 = (1 + a_3)/2$ is uniformly distributed on $[0, 1]$, and $5 \int_0^1 s^{4 - k} (1 - s)^k \dd s = \binom 4k^{-1}$. \emph{The bond projection.} For $0 \le j \le 8$, let \begin{equation*} \psi_j = \sum_{k + l = j} \binom 4k \binom 4l \, \varphi_k \otimes \varphi_l \in V \otimes V . \end{equation*} The vector $\psi_0 = \varphi_0 \otimes \varphi_0$ is an extremal-weight vector in the subspace of total spin $4$, and the operator $S^- = z_{x, 2} \partial_{z_{x, 1}} + z_{y, 2} \partial_{z_{y, 1}}$ of the bond $\{x, y\}$ satisfies $(S^-)^j \psi_0 = j! \, \psi_j$. The subspace of total spin $4$ has dimension $9$. It is invariant under $S^-$ (the action of $\mathrm{SU}(2)$ by linear substitutions extends to $\mathrm{SL}(2, \C)$, and $S^-$ is the derivative of the substitutions $z \mapsto (z_1 + t z_2, z_2)$ at $t = 0$). The vectors $\psi_0, \dots, \psi_8$ are nonzero and have distinct weights under diagonal matrices, and they span this subspace. Using \eqref{e:explicit-gram} and Vandermonde's identity, we find \begin{equation*} \|\psi_j\|^2 = \sum_{k + l = j} \binom 4k \binom 4l = \binom 8j, \qquad \langle \psi_{k + l}, \varphi_k \otimes \varphi_l \rangle = 1 . \end{equation*} The vectors $\psi_j$ are pairwise orthogonal by \eqref{e:explicit-gram}, since they involve disjoint sets of the monomials $\varphi_k \otimes \varphi_l$. Therefore the projection $P_e$ onto total spin $4$ of the bond $e = \{x, y\}$ satisfies \begin{equation}\label{e:explicit-Pe} P_e (\varphi_k \otimes \varphi_l) = \binom{8}{k + l}^{-1} \psi_{k + l} = \sum_{i + j = k + l} \frac{\binom 4i \binom 4j}{\binom{8}{k + l}} \, \varphi_i \otimes \varphi_j . \end{equation} Every coefficient in \eqref{e:explicit-Pe} has a reduced denominator that divides $70$, so that $70 P_e$ has integer entries in the monomial basis. \emph{The characteristic polynomial.} The operator $70 h_b$ is a sum of $m_b$ terms $70 P_e$, each tensored with the identity on the other sites. In particular, $70 h_b$ has integer entries in the tensor monomial basis, and its characteristic polynomial has integer coefficients. The eigenvalues of $h_b$ lie in $[0, m_b]$, since $h_b$ is a sum of $m_b$ orthogonal projections. Let $r \ge 1$ be the rank of $h_b$, and let $t_1 \le \dots \le t_r$ denote the positive eigenvalues of $70 h_b$, counted with multiplicity, so that $t_1 = 70 g_b$ and $t_r \le 70 m_b$. Since $h_b$ is self-adjoint, we have \begin{equation*} \det(\lambda - 70 h_b) = \lambda^{\dim_b - r} \prod_{i = 1}^r (\lambda - t_i) , \end{equation*} and the coefficient of $\lambda^{\dim_b - r}$ is $(-1)^r t_1 \cdots t_r$. This coefficient is a nonzero integer, so that $t_1 \cdots t_r \ge 1$. We conclude that \begin{equation*} g_b = \frac{t_1}{70} \ge \frac{1}{70 \, t_2 \cdots t_r} \ge \frac{1}{70 (70 m_b)^{r - 1}} \ge (70 m_b)^{-r} \ge (70 m_b)^{-\dim_b} . \end{equation*} This is \eqref{e:patchgapexplicit}, since $70 m_b = 140 b (b + 1)$ and $\dim_b = 5^{b^2 + 4b}$. \end{proof} \section{The constants of the main theorem}\label{s:explicit-main} We now derive \eqref{e:mainconstants}. Let $M$ and $b$ be as in \eqref{e:explicitM}. By \eqref{e:explicit-blockgap}, this value of $M$ may be used in the proof of \zcref{t:main}, and we have $L_0 = b + 4 = 2 \lceil \exp(10^{40}) \rceil + 5$. Let $L, N \ge L_0$. Combining \eqref{e:proofmain-gap} with \zcref{l:patchgapexplicit}, we obtain \begin{equation}\label{e:explicit-gapstrong} \gap(H_{L, N}) \ge \frac{g_b}{4} \ge \frac{1}{4 (70 m_b)^{\dim_b}} . \end{equation} We now simplify the right side. Since $M \le \exp(10^{40}) + 1 \le 2 \exp(10^{40})$, we have $b \le 5 \exp(10^{40}) \le \exp(2 \cdot 10^{40})$. Therefore \begin{equation*} \begin{gathered} n_b \le 2 b^2 \le \exp(5 \cdot 10^{40}), \qquad 70 m_b \le 280 b^2 \le \exp(6 \cdot 10^{40}), \\ \dim_b = \exp(n_b \log 5) \le \exp(\exp(6 \cdot 10^{40})) . \end{gathered} \end{equation*} It follows that \begin{equation*} \begin{aligned} \log \bigl( 4 (70 m_b)^{\dim_b} \bigr) &\le \log 4 + 6 \cdot 10^{40} \exp(\exp(6 \cdot 10^{40})) \\ &\le \exp(\exp(7 \cdot 10^{40})) \le \exp(\exp(10^{41})) . \end{aligned} \end{equation*} For the second inequality, we took logarithms and used that $\log(7 \cdot 10^{40}) + \exp(6 \cdot 10^{40}) \le \exp(7 \cdot 10^{40})$. Inserting this bound into \eqref{e:explicit-gapstrong}, we conclude that \begin{equation}\label{e:explicitgap} \gap(H_{L, N}) \ge \exp[-\exp(\exp(10^{41}))] = c \qquad \text{for all } L, N \ge L_0 , \end{equation} which is \eqref{e:mainconstants}. \clearpage \section{Constants and their provenance}\label{s:explicit-table} The following table lists the constants of the paper, their values, and where they are defined or verified. \begin{center} \small \begin{tabular}{>{\raggedright\arraybackslash}p{0.20\textwidth}>{\raggedright\arraybackslash}p{0.30\textwidth}>{\raggedright\arraybackslash}p{0.42\textwidth}} \toprule constant & value & provenance \\ \midrule $\varepsilon_K$ & $(16K)^{-4K}/2$ & overlap bound for sets of size at most $K$, \eqref{e:overlap} and \zcref{l:overlap} \\ $d_0$ & $24$ & width of the ends of a shell, \eqref{e:shells} \\ $M_{\mathrm g}$ & $73^4$ & threshold of the shell and cut geometry, \zcref{l:shellgeom} and \zcref{s:cut-geometry} \\ $C_\star$ & $\log 8192 < 10$ & cost of the first removal, \zcref{l:removal} \\ $3/5$, $1/2$, $1/3$ & exact & conditional means and fields of corners with one present bond, \zcref{l:threefifths,l:leaf} \\ $116/125$ & exact & lower bound for the pinned expectation of $I_{\mathrm{next}}$ in the proof of \zcref{l:response} \\ $\bar\alpha$, $\bar\beta$ & $1/6$, $9/25$ & response coefficient bounds, \zcref{l:response} \\ $\cs$, $\ct$ & $961/1000$, $187/200$ & integer certificates for all $7{,}962$ patterns, \zcref{t:H1} \\ $\gamma_{\mathrm o}$, $C_{\mathrm o}$ & $2499/2500$, $10$ & integer vector on the histories of memory $6$, \zcref{p:H2} \\ $h$ & $(1, 5/4)$ & type weights, \zcref{s:contraction} \\ $K_\star$ & $0.92320$ & bounds for $K_\tcoll(1)$ and $K_\tdiag(1)$, \zcref{t:inputs,p:tail} \\ $R$, $\epsilon_{\mathrm{out}}$ & $70$, at most $4.6 \cdot 10^{-6}$ & box radius and constant of \zcref{l:box} in the proof of \zcref{p:tail} \\ $\theta$, $A$ & $1 - 10^{-20}$, $10^{13}$ & constants of Hypothesis~(P), \eqref{e:explicitA} \\ $K_\theta$ & at most $0.9233$ & \eqref{e:explicit-Ktheta} \\ $C_{\mathrm{pin}}$ & $2 C_{\mathrm o}/(1 - \gamma_{\mathrm o}) = 5 \cdot 10^4$ & \zcref{s:contraction} \\ $C_{\mathrm{cut}}$ & less than $10^{36}$ & bound for the cut cost, \eqref{e:Ccut} and \eqref{e:explicit-Ccut} \\ $p_0$, $M_\star$ & $\exp(-2 \cdot 10^{36})$, $10^{180}$ & constants of Hypothesis~(W), \eqref{e:explicitW} \\ $\delta$ & $p_0 \varepsilon_{192} \ge \exp(-5 \cdot 10^{36})$ & success probability of one coupling trial, \eqref{e:explicitdelta} \\ $J$, $\Delta_M$ & $\lfloor M^{1/8} \rfloor$, at most $M/4$ & number of coupling trials and discrepancy bound, \eqref{e:discrepancy} and \eqref{e:explicit-Delta} \\ $M$, $b$ & $\lceil \exp(10^{40}) \rceil$, $2M + 1$ & \eqref{e:explicitM} \\ $g_b$ & at least $[140 b (b + 1)]^{-5^{b^2 + 4b}}$ & integer characteristic polynomial of the patch, \zcref{l:patchgapexplicit} \\ $L_0$, $c$ & $2 \lceil \exp(10^{40}) \rceil + 5$, $\exp[-\exp(\exp(10^{41}))]$ & \zcref{t:main}, \eqref{e:explicitgap} \\ \bottomrule \end{tabular} \end{center}