---
title: Fredkin Spin Chain
url: https://www.emergentmind.com/topics/fredkin-chain-f0f6f9f0-062b-434f-9992-bd8ece174dde
type: topic
---

# Fredkin Spin Chain

The Fredkin chain is a one-dimensional spin-\(\tfrac12\) quantum chain with interaction of three nearest neighbors, introduced as a frustration-free model whose Hamiltonian can be expressed in terms of Fredkin gates and whose exact ground-state structure is organized by Dyck paths rather than Motzkin paths [1605.03842]. In its open-boundary form, the model has an exactly solvable Dyck-path ground state; in deformed, colored, stochastic, and periodic variants it exhibits area-weighted path ensembles, anomalous entanglement scaling, polynomially or exponentially small gaps depending on regime, slow constrained dynamics, and, for periodic boundary conditions, a hidden Lie-algebraic symmetry acting on a degenerate ground-state manifold [2507.18291].

## 1. Definition and local interaction structure

In the basic spin-\(\tfrac12\) formulation, each site carries basis states \(|\uparrow\rangle\) and \(|\downarrow\rangle\). The standard open-chain bulk Hamiltonian is written as
\[
H_{\rm bulk}=\sum_{i=2}^{N-1} H_i,
\qquad
H_i = U_{i-1}P_{i,i+1}+P_{i-1,i}D_{i+1},
\]
with
\[
U_i=\tfrac12(\mathbb 1+\sigma_i^z),\qquad D_i=\tfrac12(\mathbb 1-\sigma_i^z),
\]
and
\[
P_{i,i+1}=\frac14(\mathbb 1-\vec\sigma_i\cdot \vec\sigma_{i+1}),
\]
the two-site singlet projector. The open chain is completed by boundary fields
\[
H_{\rm boundary}=|\downarrow_1\rangle\langle\downarrow_1|+|\uparrow_N\rangle\langle\uparrow_N|,
\]
which pin the left edge up and the right edge down [1605.03842, 2008.09736].

The same interaction admits a Fredkin-gate representation. If \(F_{i,j,k}\) is the controlled-SWAP gate acting with control on site \(i\), then one local term can be rewritten as
\[
H_j=(1-F_{j,j+1,j+2})+\bigl(1-\sigma^x_{j+2}F_{j+2,j+1,j}\sigma^x_{j+2}\bigr),
\]
making explicit the relation between the chain and reversible controlled-swap logic [1605.03842].

A closely related periodic formulation uses the three-site density
\[
\mathrm{F}_{j,j+1,j+2}=n_j^\uparrow\, \Pi_{j+1,j+2} +\Pi_{j,j+1}\, n^\downarrow_{j+2},
\]
where
\[
n^\uparrow=\frac12(1+\sigma^z),\qquad n^\downarrow=\frac12(1-\sigma^z),\qquad
\Pi_{j,j+1}=\frac12(1-\mathrm P_{j,j+1}),
\]
and
\[
\mathrm H=\sum_{j=1}^N \mathrm F_{j,j+1,j+2},
\]
with site labels understood modulo \(N\) [2507.18291]. What distinguishes the Fredkin chain from standard nearest-neighbor chains is precisely this conditional three-site structure: a spin-0 component of two adjacent spins interacts with a third neighboring spin [2507.18291].

## 2. Dyck paths, Fredkin moves, and exact ground states

The combinatorial description is obtained by identifying \(|\uparrow\rangle\) with an up-step and \(|\downarrow\rangle\) with a down-step. Equivalently, one may identify
\[
|\uparrow\rangle \leftrightarrow (,\qquad |\downarrow\rangle \leftrightarrow ),
\]
so that spin strings become parenthesis words and lattice walks. A Dyck path is then a walk from \((0,0)\) to \((2n,0)\) that never goes below the axis; the corresponding spin words are balanced parenthesis strings with nonnegative prefix height [1605.03842].

The local equivalence relation is generated by the Fredkin moves
\[
()) \leftrightarrow )(),\qquad (() \leftrightarrow ()(.
\]
These moves allow an adjacent matched pair \(()\) to move locally through neighboring unmatched parentheses. Basis states therefore decompose into equivalence classes \(C_{a,b}(N)\), and the open-chain boundary terms select the Dyck class \(C_{0,0}(2n)\) [1605.03842].

For even open chains, the exact ground state is the equal-weight superposition of all Dyck words,
\[
|D_n\rangle=\sum_{w\in \mathrm{Dyck}(2n)} |w\rangle,
\qquad
|D_n\rangle_{\rm norm}
=
\frac{1}{\sqrt{\mathrm{Cat}(n)}}
\sum_{w\in \mathrm{Dyck}(2n)} |w\rangle,
\]
with \(\mathrm{Cat}(n)=\frac{1}{n+1}\binom{2n}{n}\) [1605.03842]. The model is frustration free because each local projector annihilates this state separately.

The same construction extends to colored higher-spin chains. With local Hilbert space \(\mathbb C^2\otimes \mathbb C^k\), one obtains properly colored Dyck paths, in which every matched up/down pair carries the same color, and the unique ground state becomes the equal-weight sum over all such properly colored Dyck paths. In this formulation the paper constructs an \(SU(k)\)-symmetric model, and the leading term of the entanglement entropy is proportional to the square root of the length of the lattice [1605.03842].

The Dyck state also admits an exact MPS-like representation. Using infinite matrices
\[
A^{(}_{ij}=\delta_{i+1,j},\qquad A^{)}_{ij}=\delta_{i,j+1},
\qquad
v_i=\delta_{i,1},
\]
the amplitude of a word \(k_1\cdots k_N\) is \(v^T A^{k_1}\cdots A^{k_N}v\), which is nonzero exactly on Dyck words [1605.03842].

## 3. Deformations, colors, and entanglement structure

A major generalization is the deformed Fredkin chain, in which the equal-weight Dyck superposition is replaced by an area-weighted one. In the uniform colored case the Hamiltonian is
\[
H(s,t)=H_F(s,t)+H_X(s)+H_\partial(s),
\]
and the exact ground state is
\[
|GS\rangle = \frac{1}{\mathcal N}
\sum_{w\in\{\text{\(s\)-colored Dyck walks}\}}
t^{\frac12\mathcal A(w)}\,|w\rangle,
\]
where \(\mathcal A(w)\) is the area under the Dyck path \(w\) [1611.04983, 1702.03581]. At \(t=1\) this reduces to the undeformed Fredkin chain; for \(t>1\), high-area paths dominate; for \(t<1\), low-area paths dominate.

The deformation can be made site dependent while preserving frustration freeness. In the multi-parameter family, the local amplitudes \(t_{A,j}\), \(t_{B,j}\) must satisfy
\[
t_j^A=t_{j+1}^B,
\]
and the ground state is again an exact weighted superposition over colored Dyck paths [1702.03581]. A two-parameter reformulation writes the active two-site operator as
\[
\tilde P_{i,j}(\eta,\gamma)=
\frac14(\mathbb 1-\sigma_i^z\sigma_j^z)
+\frac{\eta}{4}(\sigma_i^z-\sigma_j^z)
-\frac{\gamma}{2}(\sigma_i^+\sigma_j^-+\sigma_i^-\sigma_j^+),
\]
so that the \(t\)-deformed family lies on the unit circle
\[
\eta^2+\gamma^2=1,
\]
with the original Fredkin point at \((\eta,\gamma)=(0,1)\) [2008.09736].

The entanglement structure depends sharply on \(t\) and on the number of colors. For the uniform model, the half-chain entropy is bounded for \(0<t<1\). At \(t=1\), the known undeformed results are logarithmic scaling for \(s=1\) and \(\sqrt n\) scaling for \(s>1\). For \(t>1\), the colored chain has
\[
S_n>n\log s + C(s,t),
\]
so the entropy is linear in the half-chain length, while the colorless chain remains bounded [1702.03581]. In the earlier deformation paper this same regime is described as a transition to a rainbow-like extensively entangled phase, with
\[
S_n=n\log s+O(1)
\]
for \(s>1\) and bounded entropy for \(s=1\) [1611.04983].

For \(t>1\), the ground-state magnetization is purely along \(z\),
\[
\langle s^x\rangle=\langle s^y\rangle=0,
\]
and \(\langle s^z\rangle\) exhibits a domain-wall structure [1701.00346]. In parallel, bulk tensor networks for colorful area-deformed Fredkin states can be built exactly on an upside-down step pyramid, using \(O(n^2)\) identical four-index tensors; in the Fredkin specialization the flat-step tiles of the Motzkin construction are removed, leaving only the Dyck-arch sector [1811.11974].

## 4. Gap structure, correlations, and slow dynamics

The original open Fredkin chain is gapless with a polynomially small gap. By mapping the Hamiltonian in the balanced Dyck sector to a reversible Markov chain on colored Dyck paths, the energy of the first excited state above the ground state is bounded from below by \(O(n^{-15/2})\), while a variational twisted-area state gives the upper bound
\[
\Delta(H)\le O(n^{-2}),
\]
so the gap is polynomially small in \(n\) [1609.09160].

In the deformed regime \(t>1\), variational arguments yield much smaller upper bounds. For the colored chain,
\[
\Delta \le \frac{2(4s)^n}{1+t^2}\,t^{-n^2/2},
\]
while for the colorless chain,
\[
\Delta \le t^{-n+1}.
\]
The latter provides an example of bounded entanglement entropy together with a vanishing spectral gap [1702.03581]. A related analysis of the same regime proves that the finite-size gap decays at least exponentially with system size,
\[
\Delta E \le O(t^{-2n}),
\]
and develops general bounds for von Neumann and Rényi entropies and for the entanglement spectrum for arbitrary bipartitions [1701.00346].

Static correlations are strongly anisotropic in the colorless chain. The connected longitudinal correlator satisfies
\[
\lim_{L\rightarrow \infty}\langle S_Z(2)S_Z(L-1)\rangle_c=0,
\]
whereas the transverse one remains finite,
\[
\lim_{L\rightarrow \infty}\langle S_X(2)S_X(L-1)\rangle_c=\frac{1}{32},
\]
and deep in the bulk
\[
\langle S_X(i)S_X(j)\rangle \to \frac18.
\]
A local \(S_Z\) quench therefore produces a normal light cone in \(S_Z\)-based observables but no light cone in the transverse channel [1906.09515].

The dynamical exponent is unusually large. Direct projector Monte Carlo for the lowest \(S^z_{\rm tot}=1\) excitation gives
\[
z\doteq 3.16(1),
\]
with a consistent independent estimate
\[
z=3.17(2),
\]
and interprets the excitation as an excited bond executing a constrained random walk through a Dyck background [2011.07110]. A rigorous hydrodynamic bound identifies a pseudolocal conserved matched-parentheses charge and proves
\[
\Delta(L)\le \mathcal O(L^{-5/2}),
\qquad
z\ge \frac52,
\]
establishing subdiffusion in the Fredkin dynamical universality class [2407.11110].

For the deformed Fredkin critical point at \(g_c=1\), the gap scales with critical exponents
\[
z\approx 2.69,\qquad \nu=\frac23
\]
in the \(S_z=0\) sector. Standard Kibble–Zurek scaling would give \(\gamma_0\approx 0.24\), but because the overlaps \(\langle E_l|\partial_g E_0\rangle\) scale with level-dependent exponents, the paper derives and numerically confirms the anomalous law
\[
n_{ex}\sim \tau^{-\gamma}(\ln\tau)^{(\gamma-1)/2},
\qquad
\gamma\approx 0.136
\]
instead [2207.07876].

## 5. Periodic chain, degeneracy, and hidden Lie symmetry

With periodic boundary conditions, the boundary constraints are removed, the local Fredkin moves acquire periodic extensions, and the ground state becomes highly degenerate. The ground states are organized by equivalence classes \(C_{a,b}(N)\) of Dyck paths, where a path starts at \((a,0)\), ends at \((b,N)\), has \(\Delta x=1\), \(\Delta y=\pm1\), and must hit the \(x\)-axis at least once. The corresponding class state is
\[
{C_{a,b}(N)}=
\sum_{\ell_1\ell_2\dots\ell_N\in C_{a,b}(N)}
{\ell_1\ell_2\dots\ell_N},
\]
and every vector in \(C_{a,b}(N)\) has
\[
S^z=\frac12(b-a).
\]
For each \(S^z\) there is exactly one zero-energy ground state with cyclic eigenvalue \(C=1\),
\[
v_{S^z}= \sum_{\substack{a,b\geq 0\\ b-a =2S^z}} {C_{a,b}(N)}.
\]
When \(N=2n\) is even and \(S^z=0\), there is one additional zero-energy ground state with \(C=-1\),
\[
v_0^{-}= \sum_{k=0}^{n}(-1)^k {C_{k,k}(2n)}.
\]
Hence the ground-state degeneracy is \(N+1\) for odd \(N\) and \(N+2\) for even \(N\) [2507.18291].

A central result in the periodic case is the construction of two nonlocal operators
\[
\Sigma^\pm= \sum_{\substack{r_1,\dots,r_N\in \{-1,0,1\}\\ r_1+\dots+r_N=\pm 1} }
\sigma_1^{r_1}\cdots \sigma_N^{r_N},
\]
equivalently
\[
\Sigma^\pm=\operatorname*{res}_{\lambda=0}\,
\prod_{i=1}^N \Bigl(\lambda^{-1} \sigma_i^{\pm} +1 +\lambda\sigma_i^{\mp}\Bigr),
\]
which commute with the Hamiltonian,
\[
[\Sigma^\pm,\mathrm H]=0,
\]
and act as raising and lowering operators on the cyclic-invariant ground states,
\[
\Sigma^\pm v_{S^z}=c_\pm(S^z)\,v_{S^z\pm1},
\]
away from the extremal \(S^z\) sectors [2507.18291]. More generally, the paper proves
\[
[\Sigma^Q,\mathrm F_{i,i+1,i+2}]=0
\]
for a family of charge-\(Q\) operators \(\Sigma^Q\).

From repeated commutators of \(\Sigma^\pm\), the author conjectures that the symmetry algebra generated on the ground-state manifold is
\[
\mathfrak g=
\begin{cases}
B_n=\mathfrak{so}_{2n+1}, & N=2n-1,\\[1mm]
C_n=\mathfrak{sp}_{2n}, & N=2n,
\end{cases}
\qquad
n=\lceil N/2\rceil.
\]
In this description,
\[
S^z=p+\sum_{i=1}^{\lceil N/2\rceil}\alpha_i h_i,
\]
where \(p\) is central. In the even-\(N\) case this formula exactly coincides with the analogous formula previously conjectured for the periodic Motzkin chain [2507.18291].

For even \(N\), the periodic chain also possesses
\[
\Xi=\sum_{k=0}^{N/2} (-1)^k \sum_{\ell_1\ell_2\dots\ell_N\in C_{k,k}(N)} n_1^{\ell_1} n_2^{\ell_2}\dots n_N^{\ell_N},
\]
which commutes with \(\mathrm H\), anticommutes with the cyclic shift, and exchanges the two \(S^z=0\) parity sectors. This operator is excluded from the main Lie-algebra analysis because it does not commute with the cyclic shift, but it explains the spectral doubling in the \(S^z=0\) sector [2507.18291].

## 6. Generalizations, nonergodicity, and related directions

Several later developments place the Fredkin chain in broader settings. A two-parameter extension with Hamiltonian terms \(H_i(\eta,\gamma)\) embeds the \(t\)-deformed model on the unit circle \(\eta^2+\gamma^2=1\), separating an inner ferromagnetic region from outer antiferromagnetic and dimerized regimes; throughout most of the non-ferromagnetic parts of the phase diagram, the ground state has Dyck word form, while two outer regions favor Dyck-word mismatches and exhibit exponentially small level spacing [2008.09736].

Generalized Fredkin Hamiltonians can also display Hilbert-space fragmentation and exact many-body scars. In one family,
\[
H(q)=\frac{4}{q}\sum_{i=2}^{N-2}
\left(
P_{i-1}^{\uparrow}\ket{\Phi}\bra{\Phi}_{i,i+1}
-\ket{\Phi}\bra{\Phi}_{i,i+1}P_{i+2}^{\downarrow}
\right),
\qquad
\ket{\Phi}_{i,i+1}=\ket{\uparrow\downarrow}-q\ket{\downarrow\uparrow},
\]
the minus sign introduces destructive interference preventing spin configurations from connected and therefore fractures the Hilbert space into invariant sub-spaces; within each fractured sub-sector the authors construct an exact zero-energy eigenstate with logarithmic or area-law entanglement entropy [2102.06111].

A different deformation away from the stochastic point shows a sudden change from fast thermalization to slow metastable dynamics, connected to emergent kinetic constraints and fragmentation in an effective folded model. In that setting the authors identify non-thermal eigenstates akin to quantum many-body scars, with area-law entanglement at large system size [2403.03986]. On the classical side, a stochastic Fredkin chain with generator
\[
\mathbb W = \sum_{i=2}^{N-2} f_i\left\{ c\left[\sigma_i^+\sigma_{i+1}^--(1-n_i)n_{i+1}\right] +(1-c)\left[\sigma_i^-\sigma_{i+1}^+-n_i(1-n_{i+1})\right] \right\}
\]
inherits the same Dyck-path sector, equilibrium three-phase structure, slow dynamics, and active-inactive large-deviation transition [2202.06989].

A thin-cylinder realization in fractional quantum Hall physics produces a deformed Fredkin model for the \(\nu=\tfrac52\) Moore–Read state. After truncation and restriction to the root-state-connected Hilbert-space fragment, the effective qubit Hamiltonian becomes
\[
H_F = \sum_{i=0}^{N-3} \mathcal P_i^\uparrow P_{i+1,i+2}^{\varphi(\tau)} + \mathcal P_{i,i+1}^{\varphi(\tau)} \mathcal P_{i+2}^\downarrow,
\qquad
|\varphi(\tau)\rangle=|\uparrow\downarrow\rangle-\tau|\downarrow\uparrow\rangle,
\]
with exact ground state
\[
|\psi_0\rangle = \mathcal N^{-1} \sum_{p\in\mathcal D_N} \tau^{A(p)/2}\,|p\rangle.
\]
In the thin-cylinder regime this model reproduces the Moore–Read entanglement properties and geometric-quench response [2309.04527].

The integrability question has also been addressed directly. A rigorous analysis of local conserved charges concludes that the Fredkin spin chain, under both periodic and open boundary conditions, has no nontrivial local conserved quantities beyond the obvious trivial ones; partially truncated Fredkin chains remain nonintegrable unless all three-site terms are removed, reducing the model to the pure XXX chain [2509.04838].

Finally, the name should be distinguished from a very different classical usage. A one-dimensional elementary cellular automaton with memory can realize Fredkin logic via glider collisions on a chain of binary cells, but this is not the quantum many-body Fredkin spin chain [1803.05496].

Source: https://www.emergentmind.com/topics/fredkin-chain-f0f6f9f0-062b-434f-9992-bd8ece174dde