---
title: Coherent Permutation Operators
url: https://www.emergentmind.com/topics/coherent-permutation-operators
type: topic
---

# Coherent Permutation Operators

Coherent permutation operators are permutation-based linear operators that act without measurement on a Hilbert space and therefore preserve superposition; in the most explicit circuit-level usage, they are unitary, reversible operations that reorder amplitudes among computational basis states while preserving relative phases and entanglement [2508.21667]. The literature, however, uses closely related terminology in several distinct ways: as subsystem-permutation unitaries in multipartite quantum information, as cyclic replica permutations in entanglement calculations, as basis-permutation congruences for constructing PPT bound entangled states, and, in a different algebraic sense, through coherent configurations and orbit coherence in permutation-group theory [2309.00715][1011.4706][1602.05403][1205.4960]. The common structural element is permutation action, but the meaning of “coherent” varies between operator-level reversibility, algebraic closure, and dynamical recurrence.

## 1. Definitions and terminological scope

At the operator level, a standard permutation representation on $\mathcal{H}=(\mathbb{C}^d)^{\otimes n}$ is
\[
P_d(\pi)\lvert i_1,\ldots,i_n\rangle=\lvert i_{\pi^{-1}(1)},\ldots,i_{\pi^{-1}(n)}\rangle,
\]
so the operator permutes subsystems rather than measuring them [2309.00715]. In block-encoding constructions, the term *coherent permutation operator* is used more specifically for unitary amplitude reordering on computational basis states: “coherent denotes unitary (reversible) reordering that preserves superposition and relative phases without measuring or collapsing the quantum state” [2508.21667].

Other papers use permutation operators at different granularities. In the bound-entanglement construction of Li and Qiao, $P_{mn}$ swaps the $m$th and $n$th basis elements, and generalized operators $Q_i(c)P_{mn}$ additionally multiply one row or column by a real scalar $c\neq 0,1$ [1602.05403]. In the replica approach to entanglement entropy, the local operator $\mathcal{T}_i$ cyclically permutes replicas at site $i$,
\[
\mathcal{T}_i\lvert s_1\,s_2\,\cdots\,s_n\rangle_i=\lvert s_2\,s_3\,\cdots\,s_n\,s_1\rangle_i,
\]
and products of such operators implement the replica symmetry needed for Rényi entropies [1011.4706].

| Operator family | Defining action | Representative setting |
|---|---|---|
| $P_d(\pi)$ | Permutes $n$ subsystems | Approximate orthogonality [2309.00715] |
| $\text{A\_SWAP}$ / $\text{A\_PERMUTE}$ | Reorders amplitudes coherently | Sparse-matrix block encoding [2508.21667] |
| $\mathcal{T}_i$ | Cyclically permutes replicas at one site | Replica trick and twist fields [1011.4706] |
| $P_{mn}$, $Q_i(c)P_{mn}$ | Swaps or reweights basis elements | PPT bound-entangled states [1602.05403] |

A recurring misconception is that “coherent permutation” always denotes the same notion. The operator-theoretic usage concerns unitary action on superpositions, whereas *orbit coherence* and *coherent configurations* refer to closure properties of orbit partitions or adjacency algebras rather than to amplitude-preserving circuit operations [1205.4960][1202.0241].

## 2. Representation-theoretic structure and approximate orthogonality

Permutation operators on many-body Hilbert spaces are generally not exactly orthogonal in the normalized Hilbert-Schmidt inner product. For the subsystem-permutation representation, the Gram matrix
\[
G(n,d)_{\pi_1,\pi_2}=\langle P_d(\pi_1),P_d(\pi_2)\rangle=d^{-|\pi_1^{-1}\pi_2|},
\]
with $|\sigma|=n-c(\sigma)$ and $c(\sigma)$ the number of cycles, quantifies their overlap [2309.00715]. The same work shows that $G(n,d)$ is positive semidefinite, has trace $n!$, and is invertible iff $n\le d$; moreover, for $n^2\ll d$,
\[
\|G(n,d)-I_{n!}\|_{\infty}\le \frac{n^2}{2d}.
\]

This is stronger than pairwise small overlap. Entrywise, $\langle P_d(\pi_1),P_d(\pi_2)\rangle=d^{-|\pi_1^{-1}\pi_2|}$ is small when the permutations differ substantially, and pairwise approximate orthogonality already appears when $d\gg n$. Collectively, however, control of arbitrary linear combinations requires $d\gg n^2$; that distinction is one of the central technical points of the analysis [2309.00715].

The spectrum of $G(n,d)$ is computed via Schur–Weyl duality, and explicit bounds are given for the extremal eigenvalues:
\[
\lambda_{\min}(G(n,d))=\prod_{j=1}^{n-1}\left(1-\frac{j}{d}\right),\qquad
\lambda_{\max}(G(n,d))=\prod_{j=1}^{n-1}\left(1+\frac{j}{d}\right).
\]
This representation-theoretic control supports applications including random maximally entangled states, Boson sampling, the Eggeling–Werner multipartite data-hiding scheme, LOCC and purity-testing limitations, and random-circuit poly-design arguments [2309.00715]. A plausible implication is that coherent permutation operators become an increasingly effective surrogate basis for symmetric operator expansions as the local dimension enters the stable range $n\le d$ and then the stronger regime $d\gg n^2$.

## 3. Coherent amplitude reordering in block encoding

In gate-level quantum algorithms, coherent permutation operators are introduced as a mechanism for amplitude reordering inside sparse-matrix block encoding [2508.21667]. The target unitary has the standard block form
\[
U_A=\begin{pmatrix} A/\alpha & * \\ * & * \end{pmatrix},
\]
and the implementation problem is to place prepared amplitudes into the positions required by the nonzero structure of $A$.

The basic primitive is an amplitude swap between basis states $|a\rangle$ and $|b\rangle$ that differ in one target qubit position $t$:
\[
\psi_a|a\rangle,\ \psi_b|b\rangle\longmapsto \psi_b|a\rangle,\ \psi_a|b\rangle,
\]
implemented by a generalized multi-controlled $X$ gate $C_{|c\rangle}X_{|t\rangle}$ [2508.21667]. More general amplitude permutations are then realized as sequences of such $\text{A\_SWAP}$ operations, defining $\text{A\_PERMUTE}$ between subsets of basis states. Because the transformation is unitary, it preserves superposition and entanglement during amplitude reordering.

The same framework uses these operators to compress natural compositions of MCX gates. Shift, delete, and insert operations that would otherwise be applied basis state by basis state can be regrouped by first permuting amplitudes into hardware-friendly patterns. The placement problem is cast as a minimum-Hamming-distance bijection
\[
\min_{\phi}\sum_{x\in S_2}D_H(x,\phi(x))
\]
subject to $\phi$ being a bijection, i.e. a linear assignment problem solvable by the Hungarian method [2508.21667]. The purpose is to choose control configurations compatible with nearest-neighbor connectivity and shared controls.

The resulting benefit is explicitly described as overcoming the overhead of multi-controlled $X$ gates, amplitude reordering, and hardware connectivity, and as minimizing the number of permutations while respecting nearest-neighbor constraints [2508.21667]. The examples include shift and delete operations, zero-padding to enlarge groups to powers of two, and a structured $32\times 32$ sparse matrix. The paper states that the method can minimize the number of MCX gates from exponential, when applied individually, to logarithmic, when grouped. In this sense, coherent permutation operators are not merely relabelings; they are compilation primitives that expose latent regularity in sparse-data layouts.

## 4. Replica permutations, twist fields, and entanglement entropy

A different but closely related family of coherent permutation operators appears in the replica trick for quantum spin chains. In the replicated Hilbert space, the local operator $\mathcal{T}_i$ cyclically permutes the replicas at a fixed site, and for a region $A$ one defines
\[
\mathcal{T}_A=\prod_{i\in A}\mathcal{T}_i.
\]
Rényi entropies are then written as expectation values of $\mathcal{T}_A$ in the replicated ground state [1011.4706].

For spin-$\tfrac12$, $\mathcal{T}_i$ admits an explicit elementary-matrix representation,
\[
\mathcal{T}_i=\sum_{a_1,\ldots,a_n=1}^{2}
E_{1,i}^{a_2a_1}E_{2,i}^{a_3a_2}\cdots E_{n,i}^{a_1a_n},
\]
and also an auxiliary-space trace representation [1011.4706]. These forms make precise that the operator is local on the lattice site but nontrivial in replica space. A key exchange relation,
\[
\mathcal{T}_i\mathcal{O}_{\alpha,i}=\mathcal{O}_{\alpha-1,i}\mathcal{T}_i,
\]
shows how the cyclic permutation acts on local operator insertions.

Infinite or semi-infinite products of the $\mathcal{T}_i$ become lattice precursors of branch-point twist fields of quantum field theory [1011.4706]. In the XXZ spin chain as $\Delta\to -1^+$, this framework yields exact formulas for the reduced-density-matrix eigenvalues and entropies, including
\[
S_A^{\rm R\acute{e}nyi}(n)= -\frac{nm\log 2}{1-n}
+\frac{1}{1-n}\log\left(\sum_{k=0}^m \binom{m}{k}^n\right),
\]
with large-$m$ logarithmic scaling that is reported as not conformal [1011.4706]. Here coherence is realized as local-unitary implementation of the discrete $\mathbb{Z}_n$ replica symmetry. This suggests that coherent permutation operators can function as exact lattice realizations of symmetry defects rather than only as circuit-reordering gadgets.

## 5. Permutation-based constructions of quantum states and operators

Permutation operators also serve as constructive tools for bipartite quantum states. In the construction of bound entangled states, a PPT state $\rho$ that violates the range criterion can be transformed by
\[
\rho'=(I\otimes Q)\rho(I\otimes Q)^\dagger,
\]
where $Q=Q_i(c)P_{mn}$ with $c\neq 0,1$, to produce a new state that is again PPT and again violates the range criterion [1602.05403]. The relevant mechanism is congruence transformation together with partial-transpose properties such as $P_{mn}^{T_2}=P_{mn}$. The paper further considers acting on a single eigenvector in a spectral decomposition, and checks local-unitary invariants such as $J_s(\rho)=\mathrm{Tr}(\rho^s)$, $\Omega(\rho)_{ij}=\mathrm{Tr}(\rho_i\rho_j)$, and $X(\rho)_{ijk}=\mathrm{Tr}(\rho_i\rho_j\rho_k)$ to show non-LU-equivalence of the derived states [1602.05403].

A related algebraic construction is given by Mozrzymas, Chruściński, and Sarbicki through sets of Completely Different Permutations (CDPs), where $\sigma(i)\neq \tau(i)$ for all $i$ implies Frobenius orthogonality of the associated permutation matrices [1712.03277]. A set of $n$ CDPs induces an orthogonal direct-sum decomposition
\[
\mathcal{H}\otimes\mathcal{H}=\bigoplus_{k=1}^n H_k,\qquad
H_k=\operatorname{span}\{|e_i\rangle\otimes|e_{\sigma_k(i)}\rangle\},
\]
and thereby a class of bipartite operators
\[
P[A,\mathcal{E}_n]=\sum_{k=1}^n\sum_{i,j=1}^n
a_{ij}^{(k)}|e_i\rangle\langle e_j|\otimes
|e_{\sigma_k(i)}\rangle\langle e_{\sigma_k(j)}|.
\]
If the CDPs form an abelian group, partial transpose maps the operator to another CDP operator, which the paper identifies as useful for constructing PPT states [1712.03277].

Permutation expansions also underlie a general matrix model,
\[
A=\sum_{i=1}^K \alpha_i\Pi_i,
\]
analyzed under bit-flip and phase-flip perturbations [2412.12762]. Via Sinkhorn’s theorem, any square matrix can be written as
\[
A=D_1\left(\sum_{i=1}^K \alpha_i\Pi_i\right)D_2.
\]
When the coefficients are positive, the dominant eigenvalue is described as resilient to bit-flip errors; for mixed signs, perturbations are larger, although the numerical evidence is reported to remain small when error rates are small [2412.12762]. Gershgorin-based bounds such as
\[
\Delta R_j\le 2\sum_{i=1}^K p_i|\alpha_i|
\]
for bit flips formalize that dependence. Taken together, these lines of work show that coherent permutation operators are not restricted to basis relabeling; they also furnish structured ansätze for PPT constructions, Choi matrices, and matrix encodings for quantum numerical linear algebra.

## 6. Algebraic coherence: coherent configurations and orbit coherence

In algebraic combinatorics, “coherent” refers to closure properties rather than to unitary preservation of superposition. For permutation codes in the Kendall tau metric, coherent configurations are built from orbitals of a group action on $\Omega\times\Omega$; their adjacency matrices $A_i$ satisfy $A_1=I$, $\sum_iA_i=J$, transposition closure, and the algebra relation
\[
A_iA_j=\sum_k p_{ij}^k A_k
\]
[1202.0241]. For the Kendall tau setting, the longest permutation $w_0$ defines a permutation matrix $W$ with $W^2=I$, $W=W^T=W^{-1}$, and $W$ commuting with the orbital adjacency matrices. The corresponding centralizer algebra is
\[
\mathcal{B}=\left\{\sum_{i=1}^{d} a_i\tilde{A}_i+\sum_{i=1}^{d}a_{d+i}\tilde{A}_iW: a_i\in\mathbb{R}\right\},
\]
and linear combinations in this algebra are used to derive LP relaxations of SDP upper bounds on permutation-code size [1202.0241].

Britnell and Wildon introduced *orbit coherence* for permutation groups by considering
\[
\pi(G)=\{\pi(g)\mid g\in G\},
\]
the set of orbit partitions of group elements ordered by refinement [1205.4960]. A group is join-coherent if $\pi(G)$ is closed under join and meet-coherent if it is closed under meet. Their central theorem states that the centralizer in $\mathrm{Sym}(\Omega)$ of any permutation $g$ is meet-coherent, and, under finiteness conditions on the orbits of $g$, also join-coherent; in particular, for finite $\Omega$, the orbit partitions of the centralizer form a lattice [1205.4960].

These usages are terminologically adjacent but conceptually separate from coherent amplitude reordering. A coherent configuration is an adjacency-algebra framework for symmetry reduction, and orbit coherence is a lattice property of orbit partitions. Neither notion requires the operator-level statement that superposition and relative phase are preserved without measurement.

## 7. Circuit synthesis, hierarchy structure, and dynamical generators

Permutation unitaries occupy a sharply constrained position in quantum circuit theory. In the third level of the Clifford hierarchy, every permutation gate in $C_3$ is, up to left and right multiplication by Clifford permutations, a product of Toffoli gates in staircase form; semi-Clifford permutation gates are exactly the Clifford-sandwiched mismatch-free products of controlled-$X$ gates with bounded numbers of controls [2410.11818]. The same work proves that the smallest number of qubits supporting a non-semi-Clifford permutation in $C_3$ is $7$, while every such permutation on at most $6$ qubits is semi-Clifford [2410.11818].

With the generating set $\{\sigma,\tau,\tau^{-1}\}$, where $\sigma$ swaps $1$ and $2$ and $\tau$ is the $n$-cycle, permutation complexity in $S_n$ is quadratically bounded both above and below [2207.14102]. There exists an explicit permutation requiring at least
\[
\frac{n^2-2n-7}{4}
\]
gates, every permutation admits an implementation using at most
\[
3(n-1)^2
\]
gates, and almost all permutations have lower bound
\[
\frac{n^2}{32}-3
\]
as $n\to\infty$ [2207.14102]. This places a concrete cost on coherent realization of classical reversible permutations as quantum unitaries.

A further dynamical viewpoint asks for Hamiltonians whose exponentials are permutation matrices. For a four-spin Ising chain, the unitary permutation
\[
\hat U=\hat P_{23}\hat P_{12}\hat P_{34}=e^{-i\hat HT}
\]
admits an exact logarithm, and the BCH expansion terminates in a finite expression involving commuting composite permutation operators [2001.11909]. The same paper interprets perturbations away from the exact coefficients as generating genuine quantum superpositions from an underlying permutation dynamics [2001.11909].

Finally, on infinite-dimensional Hilbert spaces, permutation operators $T_\sigma(e_n)=e_{\sigma(n)}$ are unitary and normal, but the structure of periodic points depends on cycle decomposition: $P(T)=\mathcal H$ iff all cycles are finite and uniformly bounded; if all cycles are finite but unbounded, $P(T)$ is a proper dense subspace; if some cycles are infinite, periodic points are not dense [2501.12605]. This dynamical classification shows that even when the operator is a unitary permutation, recurrence properties can vary from total periodicity to sparse periodic structure.

Source: https://www.emergentmind.com/topics/coherent-permutation-operators