---
title: Laplacian Perfect Pair State Transfer
url: https://www.emergentmind.com/topics/laplacian-perfect-pair-state-transfer
type: topic
---

# Laplacian Perfect Pair State Transfer

Laplacian perfect pair state transfer is the exact transport, under a continuous-time quantum walk generated by a graph Laplacian, of an antisymmetric two-vertex state \(e_a-e_b\) to another antisymmetric state \(e_c-e_d\) up to a unimodular phase. It arose as a generalization of Laplacian perfect state transfer between vertices, replacing localized basis states by pair states and thereby shifting the central objects of study from vertex cospectrality to strong cospectrality of real states, arithmetic constraints on Laplacian eigenvalue support, and structural mechanisms that preserve or obstruct antisymmetric dynamics [1906.01591, 2407.14376].

## 1. Formal definition and state-space viewpoint

Let \(G\) be a graph with Laplacian matrix \(L\), and let the Laplacian transition matrix be
\[
U(t)=\exp(itL).
\]
A pair state is the vector \(e_a-e_b\), where \(e_a\) and \(e_b\) are standard basis vectors. Laplacian perfect pair state transfer from \(\{a,b\}\) to \(\{c,d\}\) at time \(\tau\) means that there exists a unimodular \(\gamma\) such that
\[
U(\tau)(e_a-e_b)=\gamma(e_c-e_d).
\]
Equivalently,
\[
\left|(e_a-e_b)^\top U(\tau)(e_c-e_d)\right|=1,
\]
up to the normalization convention used in a given paper [1906.01591, 2202.04957].

The relevant spectral object is not the vertex support of a basis vector but the Laplacian eigenvalue support of the real state \(e_a-e_b\). If
\[
L=\sum_r \theta_r E_r
\]
is the spectral decomposition, then the support consists of those \(\theta_r\) for which \(E_r(e_a-e_b)\neq 0\). Perfect pair transfer requires strong cospectrality of the two pair states: for every spectral idempotent \(E_r\),
\[
E_r(e_a-e_b)=\pm E_r(e_c-e_d).
\]
This is the pair-state analogue of the strong cospectrality condition for vertex transfer [2404.16654, 2602.08684].

A basic obstruction is that if a pair state is itself a Laplacian eigenvector, then it is fixed up to phase and does not participate in nontrivial perfect transfer. In particular, fixed pair states occur when the underlying vertices are twins; in that case the antisymmetric state rotates only by a phase under \(U(t)\) and cannot transfer perfectly to a different pair state [1906.01591, 2404.16654].

## 2. Spectral criterion and distinctive dynamical features

The general criterion for Laplacian perfect pair state transfer has the same architecture as the vertex criterion, but applied to the support of a real antisymmetric state. If \(S=\{\theta_r\}\) is the Laplacian eigenvalue support of \(e_a-e_b\), then perfect pair transfer between \(e_a-e_b\) and \(e_c-e_d\) occurs if and only if three conditions hold: the two states are strongly cospectral; the supported eigenvalues are either all integers or all quadratic integers in a common quadratic field; and the support splits into \(\Lambda^+\) and \(\Lambda^-\) according to a parity rule determined by eigenvalue differences [1906.01591, 2404.16654].

More precisely, one may choose \(\theta_0\in\Lambda^+\) and a square-free \(\Delta\) so that the supported eigenvalues lie in \(\mathbb{Q}(\sqrt{\Delta})\). Writing
\[
g=\gcd\left\{\frac{\theta_0-\theta_r}{\sqrt{\Delta}}\right\},
\]
the parity of
\[
\frac{\theta_0-\theta_r}{g\sqrt{\Delta}}
\]
determines whether \(\theta_r\) belongs to \(\Lambda^+\) or \(\Lambda^-\), and the optimal transfer time is
\[
t_0=\frac{\pi}{g\sqrt{\Delta}}.
\]
The same framework is used in later work on total graphs and Q-graphs, where impossibility proofs are obtained by showing that the required support cannot satisfy the strong-cospectral and parity constraints simultaneously [2602.08684, 2407.14376].

Pair-state dynamics also exhibits features absent from vertex transfer. Symmetry and monogamy still hold: if transfer occurs from one pair state to another at a given time, the converse holds at the same time, and a given initial pair can transfer to at most one target pair. However, pair transfer also admits a transitivity phenomenon that cannot occur in vertex state transfer: if \((a,b)\to(\alpha,\beta)\) and \((b,c)\to(\beta,\gamma)\) occur at the same time, then \((a,c)\to(\alpha,\gamma)\) occurs at that time as well [1906.01591].

For bipartite graphs there is an additional correspondence: perfect pair state transfer under the Laplacian is equivalent to perfect plus-state transfer under the unsigned Laplacian when the two states are supported on opposite color classes. This places Laplacian pair transfer within a broader real-state transfer formalism in which antisymmetric and symmetric two-vertex states are intertwined by the graph bipartition [1906.01591, 2404.16654].

## 3. Canonical graph families and classification results

The strongest early classifications concern paths and cycles, where pair transfer is much more constrained than general real-state transfer.

| Graph family | Laplacian perfect pair state transfer | Representative result |
|---|---|---|
| Paths \(P_n\) | Only \(P_3\) and \(P_4\) in the pair-state classification | [1906.01591] |
| Cycles \(C_n\) | \(C_4\) in one classification; \(C_4,C_6,C_8\) in the \(s\)-pair classification | [1906.01591], [2404.16654] |
| Complete graphs \(K_n\) | None for \(n\ge 3\) | [2404.16654] |
| Antipodal distance-regular graphs with vertex PST | None except \(C_4\) | [2404.16654] |
| Complete bipartite graphs \(K_{m,n}\) | Only specific parameter sets in the complete characterization of pair and plus transfer | [2502.08103] |

For paths, the pair-state theory identifies \(P_3\) and \(P_4\) as the only examples with Laplacian perfect pair state transfer; for \(P_3\), transfer occurs between the two edge states at time \(\pi/2\), and for \(P_4\), between the end edges at time \(\frac{\sqrt{2}\pi}{2}\) [1906.01591]. This is sharply different from vertex transfer: weighted-path results show that no weighted or unweighted path on \(n\ge 3\) vertices admits Laplacian perfect state transfer between the end vertices [1708.03283]. A plausible implication is that antisymmetric two-vertex states can remain transferable in small path geometries even when vertex-localized transport is already excluded.

On cycles, two reported classifications coexist. "Pair State Transfer" states that \(C_4\) is the only cycle exhibiting Laplacian perfect pair state transfer [1906.01591]. By contrast, "A generalization of quantum pair state transfer" states that the only cycles admitting any perfect \(s\)-pair state transfer are \(C_4\), \(C_6\), and \(C_8\), and gives explicit Laplacian pair-state instances on \(C_6\) and \(C_8\) [2404.16654]. This suggests that the precise theorem hypotheses and the state family under consideration matter in low-dimensional cycle classifications.

Complete graphs provide a clean obstruction. For \(K_n\) with \(n\ge 3\), no Laplacian perfect pair state transfer occurs because the pair state \(e_a-e_b\) is already a Laplacian eigenvector: the twin structure makes it a fixed antisymmetric mode rather than a transferable one [2404.16654]. The same paper extends this rigidity to antipodal distance-regular graphs admitting vertex perfect state transfer, concluding that there is no Laplacian perfect pair state transfer in that class unless the graph is \(C_4\) [2404.16654].

In the broader real-state framework, complete bipartite graphs admit a complete characterization of pair and plus transfer. The summary of "Perfect state transfer between real pure states" reports that Laplacian pair and plus transfer in \(K_{m,n}\) occurs only for \(K_{2,4k}\), \(K_{4,4k}\), and \(K_{2,2}\) [2502.08103].

## 4. Perturbative and local construction methods

One of the main constructive mechanisms is edge perturbation between twin vertices. If \(a\) and \(b\) are twins in \(G\), let
\[
M=(e_a-e_b)(e_a-e_b)^\top.
\]
For the edge-perturbed graph \(G+\alpha\{a,b\}\), the Laplacian becomes \(L_G+\alpha M\), and the transition matrix satisfies
\[
U_{L_{G+\alpha\{a,b\}}}(t)=U_{L_G}(t)\left(I+\frac{\exp(-2i\alpha t)-1}{2}M\right).
\]
This explicit factorization yields sufficient conditions under which pair-LPST or pair-LPGST is preserved or created after perturbation [2202.04957].

The perturbative theorems distinguish three cases. If a pre-existing transferable pair state does not involve the perturbed vertices, transfer is preserved. If it involves one perturbed vertex and one external vertex, transfer is preserved provided \(\alpha T\in\pi\mathbb{Z}\). If the initial pair state is periodic, then choosing \(2\alpha T\in \pi(2\mathbb{Z}+1)\) can force transfer from \(e_a-e_q\) to \(e_b-e_q\) at time \(T\) in the perturbed graph [2202.04957]. These results generate concrete families: \(K_n-\{a,b\}\) has pair-LPST from \(e_a-e_q\) to \(e_b-e_q\) for all \(q\notin\{a,b\}\) at time \(\pi\), and deleting a matching \(S\) from \(K_n\) yields analogous transfer for every edge of the matching [2202.04957].

A more local and modular viewpoint appears in the generalized-cluster framework for real state transfer. There, a cluster \(C\) of structurally equivalent vertices supports an embedded graph \(H\), and pair states orthogonal to the all-ones direction on \(C\) evolve as though the ambient graph were absent. For Laplacian or signless Laplacian dynamics, the construction gives
\[
U_{G(H)}(t)\begin{bmatrix}x\\0\end{bmatrix}
=
e^{i\delta t\,\ell_s^\top z}
\begin{bmatrix}
U_H(t)x\\0
\end{bmatrix},
\]
so transfer in \(H\) propagates directly to transfer in the larger graph [2505.07982]. Using \(H=C_4\), the paper constructs an infinite family of non-regular graphs of maximum valency five that exhibit perfect pair state transfer under the Laplacian, adjacency, and signless Laplacian between the same pair of states at the same time [2505.07982].

These perturbative and local constructions are complementary. Edge perturbation exploits a rank-one antisymmetric mode created by twins, whereas cluster methods isolate an antisymmetric subspace whose dynamics depends only on the induced subgraph. In both cases, the transferable object is a real state with zero sum on the active support.

## 5. Involutions, half-graphs, and product constructions

Graphs with involutions furnish a systematic reduction of pair transfer to vertex transfer. For the generalized Laplacian
\[
\mathscr{L}=\Delta+qA,
\]
the choice \(q=-1\) recovers the usual Laplacian. If \(\phi\) is a non-trivial involution, then perfect pair state transfer between
\[
\frac{1}{\sqrt{2}}(e_u-e_{\phi(u)})
\quad\text{and}\quad
\frac{1}{\sqrt{2}}(e_v-e_{\phi(v)})
\]
occurs if and only if there is vertex perfect state transfer between \(u\) and \(v\) in the involution-induced half-graph with Hamiltonian \(\mathscr{L}_-\) [2604.20700]. This block-diagonal reduction extends to strong cospectrality and pretty good transfer as well [2604.20700].

That equivalence is constructive. In the 2026 formulation, almost all simple unweighted planar graphs, and almost all simple unweighted trees, can be modified by assigning loops of weight one to exactly two vertices so that the resulting graph admits pair PST relative to \(\mathscr{L}\) [2604.20700]. A related 2025 involution-based \(q\)-Laplacian study likewise derives infinite families of trees with potentials and unicyclic graphs of maximum degree three exhibiting perfect pair state transfer, again by reducing antisymmetric states in the original graph to vertex states in a smaller half-graph [2509.20749].

Regular graph products supply another route. For the tensor product \(G\times H\), "Pair state transfer in tensor product and double cover" gives necessary and sufficient conditions for Laplacian perfect pair state transfer when one factor admits perfect state transfer or pair-LPST [2509.18858]. The resulting arithmetic conditions couple the Laplacian support of a pair state in one factor to phase data from the transfer time in the other. Explicit families include \(K_n\times P_2\), which admits pair-LPST for any \(n\ge 3\), and \(K_{2n}\times C_4\), which has pair-LPST at time \(\pi/2\) [2509.18858].

For double covers \(G\ltimes H\), the same paper characterizes pair-LPST in terms of pair periodicity or pair transfer for the matrices \(A_G+A_H\) and \(A_G-A_H\). One explicit example is \(K_n\ltimes K_n\), which admits pair-LPST at time \(\pi/2\) [2509.18858]. These product results expose a recurrent principle: exact antisymmetric transfer survives when the product decomposition preserves a controllable two-phase splitting of the relevant support.

## 6. Nonexistence theorems and the shift to pretty good transfer

The recent literature on derived regular graphs emphasizes how restrictive exact Laplacian pair transfer is. For Q-graphs of \(r\)-regular graphs, if \(r+1\) is prime or a power of \(2\), then the Q-graph \(Q(G)\) does not have Laplacian perfect pair state transfer [2407.14376]. This covers, in particular, \(Q(C_n)\) and Q-graphs of generalized Petersen graphs in the cases stated in the paper. The same work gives a sufficient condition for pair-LPGST: if \(G\) has pair-LPST and \(r+1\) is prime, then \(Q(G)\) can inherit pair-LPGST under explicit bipartite or non-bipartite hypotheses, and \(Q(C_4)\) is given as an example [2407.14376].

For total graphs \(\mathcal{T}(G)\) of \(r\)-regular graphs, the obstruction is formulated spectrally. If \(r>2\) and \(r+1\) is not a Laplacian eigenvalue of \(G\), then \(\mathcal{T}(G)\) fails to exhibit Laplacian perfect pair state transfer; if \(G=K_n\) with \(n>3\), then \(\mathcal{T}(K_n)\) also fails to exhibit Laplacian perfect pair state transfer [2602.08684]. Nevertheless, under mild arithmetic conditions, \(\mathcal{T}(G)\) exhibits Laplacian pretty good pair state transfer, and the paper derives infinitely many such total graphs from cocktail party graphs and hypercubes [2602.08684].

| Derived graph class | Exact pair-LPST | Approximate pair transfer |
|---|---|---|
| Q-graph \(Q(G)\) of an \(r\)-regular graph | No if \(r+1\) is prime or a power of \(2\) | Pair-LPGST under stated inheritance conditions |
| Total graph \(\mathcal{T}(G)\) of an \(r\)-regular graph | No if \(r>2\) and \(r+1\notin \operatorname{Spec}_L(G)\) | Pair-LPGST under mild arithmetic conditions |

A consistent pattern emerges. Exact Laplacian pair state transfer is governed by strong cospectrality together with stringent integrality or quadratic-integrality conditions on the supported spectrum. Derived graph operations often introduce square roots or incompatible parity data, destroying those conditions even when the original graph has exact transfer. By contrast, pretty good pair state transfer can persist because Kronecker-type approximation arguments allow quasi-periodic phases to align arbitrarily well, even when exact commensurability fails [2407.14376, 2602.08684].

In that sense, Laplacian perfect pair state transfer occupies a narrow but structurally rich regime: rigid enough to admit sharp classifications and no-go theorems, yet flexible enough to support constructive theories based on twin perturbations, involutions, clusters, and graph products.

Source: https://www.emergentmind.com/topics/laplacian-perfect-pair-state-transfer