---
title: Fredkin Spin Chain
url: https://www.emergentmind.com/topics/fredkin-spin-chain
type: topic
---

# Fredkin Spin Chain

The Fredkin spin chain is a one-dimensional spin-$\tfrac12$ quantum lattice model with genuine three-site interactions, most commonly defined with open boundaries that pin the first spin up and the last spin down, and whose frustration-free ground state is exactly expressible in terms of Dyck-path combinatorics. In its original form the Hamiltonian can be written through controlled-SWAP (Fredkin) gates; in colored and area-deformed variants, the same combinatorial structure yields logarithmic, square-root, or volume-law entanglement, polynomially or faster closing gaps, anomalous transport, and a variety of nonergodic phenomena [1605.03842] [1702.03581].

## 1. Hamiltonian and local constraint structure

A standard open-chain formulation on $N$ sites uses the Fredkin gate
\[
F_{i,j,k}
=|\downarrow\rangle\langle\downarrow|_i\otimes I_{j,k}
+|\uparrow\rangle\langle\uparrow|_i\otimes \mathrm{SWAP}_{j,k},
\]
with boundary projectors fixing the left edge to $\uparrow$ and the right edge to $\downarrow$. One convenient expression is
\[
H = H_{\partial}+\sum_{j=1}^{N-2}H_j,\qquad
H_{\partial}=P_1^-+P_N^+,
\]
\[
H_j=(1-F_{j,j+1,j+2})+(1-\sigma^x_{j+2}F_{j+2,j+1,j}\sigma^x_{j+2}),
\]
where $P_j^\pm=(1\pm \sigma_j^z)/2$ [1605.03842]. Equivalent projector formulations use singlet projectors on nearest-neighbor bonds conditioned by an adjacent spin, for example
\[
F_i=U_{i-1}P_{i,i+1}+P_{i-1,i}D_{i+1},
\]
with $U_i=\tfrac12(1+\sigma_i^z)$, $D_i=\tfrac12(1-\sigma_i^z)$, and $P_{i,i+1}=\tfrac14(1-\boldsymbol\sigma_i\!\cdot\!\boldsymbol\sigma_{i+1})$ [1805.00532].

The local moves implemented by these projectors are the Fredkin moves: adjacent $\uparrow\downarrow$ pairs can “bubble” past a neighboring spin only when the relevant control spin is in the correct state. This local kinetic rule is the quantum counterpart of a constrained nearest-neighbor exchange process. The Hamiltonian is frustration-free: each local projector annihilates the exact ground state [1906.09515].

The colored generalization enlarges the on-site basis to
\[
\{\ket{u^1},\ldots,\ket{u^s};\ket{d^1},\ldots,\ket{d^s}\},
\]
or equivalently $\{\ket{\uparrow^c},\ket{\downarrow^c}\}_{c=1}^s$, and adds nearest-neighbor color-matching terms that force each matched up-step and down-step to carry the same color [1702.03581]. In the original SU$(k)$ formulation, the higher-spin half-integer extensions are realized as properly colored Dyck-path models with $k$ colors [1605.03842].

## 2. Dyck-path ground states and exact representations

The defining combinatorial map identifies $\uparrow$ with an up-step and $\downarrow$ with a down-step. A basis state is therefore a lattice walk, and the open-boundary ground space selects Dyck paths: walks from $(0,0)$ to $(2n,0)$ that never cross below the horizontal axis. For the uncolored chain of length $2n$, the unique ground state is the equal-weight Catalan superposition
\[
|D_n\rangle
=\frac{1}{\sqrt{C_n}}\sum_{w\in\mathrm{Dyck}_{2n}}|w\rangle,
\qquad
C_n=\frac1{n+1}\binom{2n}{n},
\]
while the colored version is
\[
\bigl|\mathcal D_{2n}^s\bigr\rangle
=\frac1{\sqrt{s^nC_n}}
\sum_{\substack{\text{$s$-colored}\\ \text{Dyck paths }w}}
|w\rangle
\]
[1605.03842] [1609.09160].

Area-deformed Fredkin chains reweight the Dyck configurations by the area $\mathcal A(w)$ under the path. In the uniform deformation, one exact ground-state convention is
\[
|\Psi_0\rangle\propto \sum_w t^{\frac12\mathcal A(w)}|w\rangle,
\]
while another uses $t^{\mathcal A(w)}$; both are explicit weighted Dyck-path superpositions differing by parameterization conventions [1702.03581] [1811.11974]. More generally, the multiparameter deformation assigns position-dependent weights while preserving frustration-freeness through the constraint $t_{B,j}=t_{A,j+1}$ [1702.03581].

These exact states admit explicit tensor-network descriptions. The original uncolored model has an exact matrix-product-state representation built from shift matrices $A^\uparrow$ and $A^\downarrow$ satisfying
\[
A^\uparrow A^\downarrow=I,\qquad
A^\downarrow A^\uparrow=I-|1\rangle\langle 1|,
\]
with an exact truncation of the bond dimension to $2n$ for a chain of length $2n$ [1605.03842]. For the area-deformed “rainbow” states, an exact two-dimensional bulk tensor network on an inverted step pyramid of identical four-leg tensors $B(t)$ reproduces the weighted Dyck-path amplitudes, with each horizontal bulk arrow contributing a factor $\sqrt t$; the construction uses only $O(n^2)$ copies of a single tensor, but the resulting network is not compressible into an MPS of finite bond dimension [1811.11974].

## 3. Entanglement structure, correlations, and phase organization

The Fredkin chain is notable for entanglement scaling that goes well beyond the standard one-dimensional area-law paradigm. In the undeformed uncolored model, the half-chain entanglement entropy grows logarithmically:
\[
S_{1/2}(n)=\tfrac12\log n+O(1),
\]
or more explicitly
\[
S_{F,A}
=\tfrac12\ln n+\tfrac12\ln\frac{\pi}{4}+\gamma-\tfrac12+o(1)
\]
[1605.03842] [1806.04049]. In the undeformed colored model, the von Neumann entropy scales as
\[
S_{F,A}
=(2\ln s)\sqrt{\frac n\pi}+\frac12\ln n+O(1),
\]
giving a square-root violation of the area law [1806.04049].

For the Rényi entropies of the colored undeformed chain, the asymptotics are non-analytic at $\alpha=1$: for $0<\alpha<1$ they obey a volume law, at $\alpha=1$ they scale as $\sqrt n$, and for $\alpha>1$ they scale as $\log n$ [1806.04049]. This establishes an entanglement transition in Rényi index rather than in a microscopic coupling.

In the uniform area-deformed chain, the phase structure is controlled by the deformation parameter $t$ and the number of colors $s$ [1702.03581].

| Regime | Half-chain entanglement | Gap characterization |
|---|---|---|
| $t<1$, any $s\ge1$ | $S_n=O(1)$ | gapped |
| $t=1$, $s=1$ | $S_n\sim \log n$ | gapless |
| $t=1$, $s>1$ | $S_n\sim \sqrt n$ | gapless |
| $t>1$, $s=1$ | bounded $S_n=O(1)$ | gapless |
| $t>1$, $s>1$ | $S_n\sim n\log s$ | gapless |

For $t>1$ and $s>1$, one proves
\[
S_n(s,t)>n\log s+O(1),
\]
and, in particular, for the half-chain length $n$,
\[
S_{1/2}=n\log s,
\]
which is a full volume law [1702.03581]. The tensor-network description of this regime is the “rainbow phase,” in which dominant configurations are concentric arcs crossing the bipartition extensively [1811.11974].

The static correlation structure is also unconventional. In the undeformed spin-$\tfrac12$ chain, the connected longitudinal correlator $\langle S_Z(i)S_Z(j)\rangle_c$ decays to zero at large separation, but the connected transverse correlator $\langle S_X(i)S_X(j)\rangle_c$ does not: at the edge,
\[
\lim_{L\to\infty}\langle S_X(2)S_X(L-1)\rangle_c=\frac1{32},
\]
and deep in the bulk the corresponding limit is $\frac18$ [1906.09515]. Accordingly, the cluster decomposition property holds along the $Z$ axis and fails in the $XY$ plane [1906.09515].

For colorful Fredkin chains, long-distance entanglement persists between disjoint subsystems. For two edge spins, the negativity is nonzero for $q\ge3$ and has the asymptotic limit
\[
\mathcal N_\infty=\frac{(q-1)(q-3)}{8q},
\]
while for $q=2$ the reduced state is PPT-entangled. The edge mutual information remains finite for all $q>1$, whereas in the colorless case $q=1$ it vanishes [1904.05205]. This is consistent with the failure of cluster decomposition in transverse observables.

## 4. Spectral gap, transport, and anomalous dynamics

The original Fredkin chain is gapless in the thermodynamic limit, but its finite-size gap closes unusually slowly. For the balanced sector of the spin chain on $2n$ sites, rigorous bounds give
\[
\Omega(n^{-15/2})\le \Delta(2n)\le O(n^{-2}),
\]
hence
\[
\Delta(2n)=\Theta(n^{-c}),\qquad 2\le c\le \frac{15}{2},
\]
with the lower bound derived via a mapping to Glauber dynamics on Dyck paths and the Diaconis–Saloff-Coste comparison theorem, and the upper bound obtained from a Brownian-excursion variational state [1609.09160].

In the deformed chain, the gap can close much faster. For the uniform $t>1$, $s>1$ phase,
\[
\Delta=O\!\bigl((4s)^n\,t^{-n^2/2}\bigr),
\]
a super-exponential upper bound in $n^2$ [1702.03581]. In the colorless $t>1$, $s=1$ case one has
\[
\Delta\le t^{-n+1},
\]
which is exponentially small even though the entanglement remains bounded [1702.03581]. A related one-parameter deformation proved
\[
\Delta E(2n,t)=O(t^{-2n}),\qquad t>1,
\]
and established a domain-wall magnetization profile with $\langle s_j^x\rangle=\langle s_j^y\rangle=0$ and $\langle s_j^z\rangle\approx +\tfrac12$ on the left half, $\approx -\tfrac12$ on the right half for spin-$\tfrac12$ [1701.00346].

Direct dynamical simulations reveal slow relaxation with a large dynamical exponent. Quantum Monte Carlo and finite-size scaling studies reported
\[
z=3.16(1),
\]
interpreting the slow mode as a “canted bond” executing a constrained random walk through a Dyck-word background [2011.07110]. More recently, a pseudolocal matched-parentheses conserved charge was identified in the periodic Fredkin chain, yielding an exact variational upper bound
\[
\Delta E(L)=O(L^{-5/2})
\]
and therefore a rigorous hydrodynamic lower bound
\[
z\ge \frac52
\]
for Fredkin-constrained transport [2407.11110].

The response to local perturbations is also anisotropic. After a local $h_zS_Z(j_0)$ quench, the change in $ZZ$ connected correlators shows a conventional light cone, whereas the $XX$ connected correlators respond almost immediately at all distances, reflecting pre-existing long-range transverse correlations [1906.09515].

Adiabatic driving across the deformed critical point at $g_c=1$ gives a further anomaly. In that setting, finite-size scaling of the first gap yields
\[
E_1(N,g=1)\sim N^{-z},\qquad z\approx 2.69,
\]
with
\[
\nu\approx \frac23,
\]
and the excitation density obeys
\[
n_{\rm ex}\sim \tau^{-\gamma}(\ln\tau)^{(\gamma-1)/2},
\qquad
\gamma=\frac{\gamma_0}{2-\gamma_0},
\qquad
\gamma_0=\frac{\nu}{1+z\nu},
\]
rather than the standard Kibble–Zurek form [2207.07876].

## 5. Deformations, generalized descendants, and nonergodic sectors

Several families interpolate away from the original frustration-free point. A two-parameter generalization with projector
\[
\tilde P_{i,j}(\eta,\gamma)
=\frac14(1-\sigma_i^z\sigma_j^z)
+\frac{\eta}{4}(\sigma_i^z-\sigma_j^z)
-\frac{\gamma}{2}(\sigma_i^+\sigma_j^-+\sigma_i^-\sigma_j^+)
\]
has the exactly solvable $t$-deformed line on the unit circle
\[
r=\sqrt{\eta^2+\gamma^2}=1,\qquad
\eta=\frac{1-t^2}{1+t^2},\quad
\gamma=\frac{2t}{1+t^2},
\]
where the ground state remains an area-weighted Dyck-word superposition [2008.09736]. Outside that line, the same study identified ferromagnetic, antiferromagnetic, dimerized, and non-Dyck “cat’s-ear” regions, with exponentially small gaps in the non-Dyck domes according to exact diagonalization for small $N$ [2008.09736].

A different interpolation
\[
H(\lambda)=\sum_{i=2}^{N-1}G_i(\lambda)+H_{\rm bdy}
\]
connects the Fredkin point $\lambda=0$ to the ferromagnetic and antiferromagnetic Heisenberg chains at $\lambda=\pm1$ [1805.00532]. Its zero-temperature phase diagram contains an $XY$-FM region for $-1\le\lambda\le0$, a dimer plus Ising-AF region for $0<\lambda<\lambda_{c2}\approx0.73$, and a Heisenberg-AF regime for $\lambda_{c2}<\lambda\le1$, with a crossover between two dimer patterns at $\lambda_{c1}\approx0.295$ [1805.00532]. The Fredkin point is “brittle” in the sense that an infinitesimal $\lambda>0$ opens a gap and destroys the gapless frustration-free behavior [1805.00532].

Generalized Fredkin Hamiltonians also realize Hilbert-space fragmentation and exact scars. In one spin-$\tfrac12$ family,
\[
H(q)=\frac{4}{q}\sum_{i=2}^{N-2}
\Bigl[
P_{i-1}^\uparrow\,|\Phi\rangle\langle\Phi|_{i,i+1}
-
|\Phi\rangle\langle\Phi|_{i,i+1}\,P_{i+2}^\downarrow
\Bigr],
\]
the dynamics preserve the order of an emergent sequence of red and blue particles, so the Hilbert space fractures into disconnected Krylov subsectors labeled by that fixed color string and the hole number [2102.06111]. Each subsector contains one exact closed-form zero-energy scar state for arbitrary $q$, and the half-chain Rényi-2 entropy satisfies
\[
S_2\le \log\!\Bigl[\min\!\bigl(2n_+ + 1,\;N-2n_+ +1\bigr)\Bigr]
\]
[2102.06111].

A deformed stochastic Fredkin chain with parameters $(c,s)$ exhibits a sharp dynamical crossover between fast thermalization and slow metastable dynamics near its stochastic point $(c,s)=(\tfrac12,0)$ [2403.03986]. In the large-$s$ limit, an emergent folded Hamiltonian fragments the Hilbert space into exponentially many disconnected pieces; exact diagonalization finds a set of $N/2$ non-thermal eigenstates with low entanglement and anomalous local observables, while variational MPS shows that their maximum bipartite entanglement saturates with system size [2403.03986]. This suggests that Fredkin-type kinetic constraints provide a direct route from frustration-free combinatorics to weak ergodicity breaking.

## 6. Periodic-chain symmetries, conservation laws, and integrability status

Under periodic boundary conditions, the Fredkin chain changes qualitatively. The periodic Hamiltonian can be written as
\[
H=\sum_{j=1}^{N}\mathrm F_{j,j+1,j+2},\qquad
\mathrm F_{j,j+1,j+2}
=n_j^\uparrow\,\Pi_{j+1,j+2}+\Pi_{j,j+1}\,n_{j+2}^\downarrow,
\]
with indices understood modulo $N$ [2507.18291]. The ground state is no longer unique: the zero-energy subspace is spanned by vectors built from equivalence classes $C_{a,b}(N)$ of Dyck paths related by Fredkin moves, and its dimension is $N+1$ for odd $N$ and $N+2$ for even $N$ [2507.18291].

A notable recent development is the construction of two nonlocal operators $\Sigma^\pm$ commuting with the periodic Hamiltonian. Acting on the ground-state multiplet, they serve as raising and lowering operators in $S^z$, and the Lie algebra generated by nested commutators of $\Sigma^\pm$ closes on
\[
B_n=\mathfrak{so}(2n+1)\quad\text{for }N=2n-1,
\qquad
C_n=\mathfrak{sp}(2n)\quad\text{for }N=2n,
\]
with rank $n=\lceil N/2\rceil$ [2507.18291]. In even length, the decomposition of total $S^z$ into Cartan generators plus a central element coincides with a corresponding formula proposed for the periodic Motzkin chain [2507.18291].

These nonlocal symmetries do not imply Yang–Baxter integrability. A separate analysis showed that, under both periodic and open boundary conditions, the Fredkin spin chain has no nontrivial local conserved charges beyond the Hamiltonian itself; the same conclusion holds for truncated variants obtained by deleting subsets of the three-site interaction terms [2509.04838]. The proof proceeds by an exhaustive Pauli-string commutator analysis using a column method to exclude $k$-local charges for all $4\le k\le N/2$ under periodic boundaries and boundary-supported local charges under open boundaries [2509.04838]. The present status is therefore that the Fredkin chain is partially solvable in its ground-state sector, but nonintegrable in the local-conservation-law sense.

The coexistence of exact frustration-free structure, combinatorial solvability, anomalous entanglement, subdiffusive transport bounds, fragmentation phenomena, and nonintegrability makes the Fredkin spin chain a distinctive class of constrained quantum many-body systems rather than a conventional integrable spin chain.

Source: https://www.emergentmind.com/topics/fredkin-spin-chain