---
title: Pretty Good Fractional Revival
url: https://www.emergentmind.com/topics/pretty-good-fractional-revival
type: topic
---

# Pretty Good Fractional Revival

Pretty good fractional revival (PGFR) is the asymptotic form of fractional revival in continuous-time quantum walks on graphs. In the two-vertex setting, for a graph with adjacency matrix \(A\) and transition matrix \(H(t)\), PGFR between distinct vertices \(u\) and \(v\) means that there is a sequence of times \(\{t_k\}\) and complex numbers \(\alpha,\beta\) with \(|\alpha|^2+|\beta|^2=1\) and \(\beta\neq 0\) such that
\[
\lim_{k\to\infty} H(t_k)e_u=\alpha e_u+\beta e_v.
\]
It is the approximate counterpart of exact fractional revival, just as pretty good state transfer is the approximate counterpart of perfect state transfer. The literature also develops a subset version: for a subset \(K\), PGFR requires that the closure of the time-evolved \(K\times K\) submatrices contains a unitary that is not a scalar multiple of the identity [2005.00492], [2311.18143].

## 1. Formal definition and basic variants

The standard setting is a continuous-time quantum walk generated by a Hermitian matrix, usually the adjacency matrix \(A\), with transition matrix written in the literature as either \(U(t)=e^{-itA}\) or \(H(t)=\exp(itA)\). Exact fractional revival from \(u\) to \(v\) at time \(\tau\) means
\[
U(\tau)e_u=\alpha e_u+\beta e_v,
\]
with \(|\alpha|^2+|\beta|^2=1\) and \(\beta\neq 0\). Perfect state transfer is the special case \(\alpha=0\), and periodicity is the case \(\beta=0\) [1801.09654], [2004.01129].

PGFR weakens exact equality to arbitrarily close approximation. In the pairwise formulation, it is given by a convergent subsequence of evolved basis states. In the arbitrary-subset formulation, if \(M\) is a real symmetric matrix and \(K\subset X\), then \(M\) exhibits PGFR with respect to \(K\) if
\[
\mathrm{cl}\left\{ \exp(itM)_{K \times K} : t \geq 0 \right\} \cap \mathcal{U}(K) \supsetneq \{ p I_{K \times K} : p \in \mathbb{C} \}.
\]
Equivalently, there exists a sequence \(t_k\to\infty\) such that \(\exp(i t_k M)_{K\times K}\) converges to a \(K\times K\) unitary with at least two distinct eigenvalues [2311.18143].

For \(|K|=2\), the subset definition reduces to the familiar vertex-pair condition. One formulation states that for every \(\varepsilon>0\), there is a time \(t\) such that
\[
|\exp(itA)_{u,u}|^2+|\exp(itA)_{u,v}|^2>1-\varepsilon.
\]
This captures the idea that an initial state localized at \(u\) is asymptotically confined to the span of \(e_u\) and \(e_v\) [2311.18143].

## 2. Spectral and algebraic framework

The spectral theory of PGFR is built from the same decomposition used for exact fractional revival. If
\[
A=\sum_r \theta_r E_r,
\qquad
U(t)=\sum_r e^{-it\theta_r}E_r,
\]
then the phenomenon is controlled by how the spectral idempotents \(E_r\) act on the distinguished vertices or subset [2004.01129].

For two vertices, the key notions are fractional cospectrality and strong fractional cospectrality. The pairwise theory initiated in “Approximate quantum fractional revival in paths and cycles” shows that PGFR occurs between \(u\) and \(v\) if and only if \(u\) and \(v\) are fractionally cospectral and a number-theoretic condition holds on the eigenvalues. If \(\Pi_1,\Pi_2\) are the two groups of eigenvalues determined by fractional cospectrality, then for any integers \(\ell_i\),
\[
\sum_{i\in \Pi_1}\ell_i\lambda_i+\sum_{j\in \Pi_2}\ell_j\lambda_j=0,
\qquad
\sum_i \ell_i=0
\]
must imply
\[
\sum_{i\in \Pi_1}\ell_i\neq \pm1.
\]
The same work proves that PGFR implies strong fractional cospectrality [2005.00492].

For arbitrary subsets, the theory is organized by the partition \(\mathcal{P}_K\) induced by spectral overlap on \(K\). A central result states that a matrix \(M\) exhibits PGFR with respect to \(K\) if and only if \(\mathcal{P}_K\) is non-degenerate. The non-degeneracy condition is expressed as simultaneous approximation of the phases \(t\theta_j\) modulo integers on each part of the partition, and a Kronecker-type lemma converts this into an integer-relation obstruction. An irreducibility criterion is also available: if the characteristic polynomial factors into irreducible blocks whose trace-to-degree ratios differ, then the partition is non-degenerate, and PGFR follows [2311.18143].

The exact theory remains relevant in the approximate setting. The framework of \(K\)-fractional revival and the associated ratio condition show how commuting spectral partitions and rational relations among eigenvalue differences govern exact revival; the PGFR criteria arise when these phase alignments are approached through Diophantine approximation rather than attained exactly [2004.01129].

## 3. Paths, cycles, and Laplacian models

In simple unweighted graphs, exact fractional revival is scarce. For adjacency dynamics, fractional revival in paths occurs if and only if \(n=2,3,4\), and in cycles if and only if \(n=4\) or \(n=6\) [1801.09654]. PGFR enlarges this landscape substantially.

For adjacency walks on paths \(P_n\), the classification is complete. PGFR occurs for symmetric pairs when \(n=p2^k-1\) for a prime \(p\) and \(k\geq 0\), with the relevant vertices \(a\) and \(p2^k-a\) and \(a\) a multiple of \(2^{k-1}\). There is also a genuinely asymmetric infinite family: if \(n=5\cdot 2^k-1\), PGFR occurs between \(2^k\) and \(3\cdot 2^k\), and by symmetry between \(2\cdot 2^k\) and \(4\cdot 2^k\). For cycles \(C_n\), PGFR occurs between antipodal vertices if and only if \(n=2p^k\), where \(p\) is an odd prime and \(k\geq 1\) [2005.00492].

Laplacian dynamics yield a different classification. In paths, Laplacian PGFR occurs only between symmetric vertices \(a\) and \(n+1-a\), and only when either \(n=p^\ell\) for a prime \(p\) and \(a\neq (p^\ell+1)/2\), or \(n=2p^\ell\) for an odd prime \(p\) with \(a=(p^\ell+1)/2\) or \(a=(3p^\ell+1)/2\). For double stars \(S(m,n)\), PGFR occurs exactly in three cases: \(n=m\) with the two non-pendant vertices, \(n\neq m\) and \(n=2\) with the two pendant neighbors of the degree-3 vertex, and \(S(1,1)=P_4\) with the two extremal vertices [2010.10465].

| Family | PGFR condition | Source |
|---|---|---|
| Paths \(P_n\) (adjacency) | \(n=p2^k-1\) for symmetric pairs; also \(n=5\cdot 2^k-1\) for specific asymmetric pairs | [2005.00492] |
| Cycles \(C_n\) (adjacency) | Antipodes iff \(n=2p^k\), \(p\) odd prime | [2005.00492] |
| Paths \(P_n\) (Laplacian) | Only symmetric pairs, with \(n=p^\ell\) or \(n=2p^\ell\) in the classified cases | [2010.10465] |
| Double stars \(S(m,n)\) (Laplacian) | Exactly the three cases in the classification theorem | [2010.10465] |

These classifications establish a recurring theme: exact fractional revival is exceptional, while PGFR survives in broader arithmetic families because Kronecker approximation replaces exact phase locking by asymptotic phase alignment.

## 4. Structural mechanisms: twins, perturbations, and subset induction

One mechanism for PGFR comes from highly constrained local symmetry. For twin vertices in a weighted graph, the exact fractional revival theory gives explicit spectral congruences in terms of a unique negative-support eigenvalue \(\theta\) and the positive-support eigenvalues \(\lambda_1,\dots,\lambda_r\). In this setting, if
\[
\frac{\lambda_1-\theta}{\lambda_1-\lambda_2}
\]
is irrational, then proper pretty good state transfer occurs; the paper states that this implies PGFR. The same work also shows a monogamy phenomenon for exact revival among twins: a twin vertex can be involved in proper FR with at most one other vertex, and only another twin [2303.04952].

A second mechanism is perturbative. The arbitrary-subset theory with magnetic fields considers a diagonal perturbation
\[
M_K=M+Q D_K,
\]
where \(Q\) is transcendental and \(D_K\) is the diagonal projection onto \(K\). The perturbed characteristic polynomial factors as
\[
\phi(M_K,t)=P_0(t)\prod_{j=1}^{|K|}P_j(t),
\]
and the minimal polynomials \(P_j\) are irreducible over the base field with \(Q\) adjoined. If two of these blocks have different trace-to-degree ratios, then the partition is non-degenerate, and \(M+Q D_K\) has PGFR with respect to \(K\). This gives a general method for inducing PGFR by constant diagonal perturbations on arbitrary subsets [2311.18143].

These results enlarge the scope of PGFR beyond fixed, highly symmetric graphs. A plausible implication is that subset-based PGFR is not merely a property to be detected in an existing network; it can also be engineered by altering spectral blocks through perturbation.

## 5. Cayley, semi-Cayley, and related group-based graphs

Cayley-type graphs provide the most extensive exact and pretty good classifications. For exact fractional revival on abelian Cayley graphs, the decisive ingredients are already rigid: the difference \(a=x-y\) must be of order two, the graph must be integral, and the eigenvalue phases must split according to the two character classes
\[
G_0=\{g:\chi_a(g)=1\}, \qquad G_1=\{g:\chi_a(g)=-1\}
\]
with
\[
e^{it\lambda_g}=\alpha+\beta \quad (g\in G_0), \qquad e^{it\lambda_g}=\alpha-\beta \quad (g\in G_1).
\]
The quasi-abelian extension replaces abelian characters by irreducible characters of a finite group and requires that \(uv^{-1}\) be a central involution; exact fractional revival again forces integrality [2208.05107], [2502.14330].

For PGFR on abelian Cayley graphs, the later criterion is explicitly Diophantine. If \(b-a\) has order \(2\), then \(\mathrm{Cay}(G,S)\) has PGFR between \(a\) and \(b\) if and only if, for all integers \(\ell_1,\dots,\ell_{n-1}\), the relation
\[
\sum_{r=1}^{n-1}\ell_r(\lambda_r-\lambda_0)=0
\]
implies
\[
\sum_{r\in X_2}\ell_r\neq \pm1,
\]
where
\[
X_1=\{r:\chi_r(b-a)=1\},\qquad X_2=\{r:\chi_r(b-a)=-1\}.
\]
This criterion yields infinite classes of circulant graphs with PGFR, infinite classes without PGFR, and a complete characterization of unitary Cayley graphs \(\mathrm{Cay}(\mathbb{Z}_n,U(n))\): for \(n\ge 4\), PGFR occurs if and only if \(n=2p\), where \(p\) is a prime [2503.20367].

The classification is sharpened further for unitary and quadratic unitary Cayley graphs. The unitary Cayley graph \(X_n\) admits PGFR if and only if \(n=2\) or \(n=2p\), where \(p\) is a prime, and \(X_n\) admits fractional revival if and only if it admits PGFR. For the quadratic unitary Cayley graph \(G_n\), PGFR occurs if and only if
\[
n\in\{2,8,2p\},
\]
where \(p\) is any prime, whereas exact fractional revival occurs only for
\[
n\in\{2,4,2p\}\quad\text{with}\quad p\equiv 3\pmod 4.
\]
This shows that in \(G_n\), unlike \(X_n\), PGFR and exact fractional revival do not coincide [2508.18068].

Non-abelian families display similarly sharp arithmetic behavior. For Cayley graphs over dicyclic groups \(T_{4n}\), the existence of PGFR is characterized by number-theoretic linear independence conditions on representation-theoretic eigenvalues. The paper gives positive results for prime powers, powers of two, and certain mixed orders, and negative results when the prime-factor structure forces the relevant relations to fail [2312.10985]. Semi-Cayley graphs over abelian groups are treated primarily in the exact setting, but the same paper states that its integrality and gcd/time arguments suggest pretty good fractional revival when integrality fails but the spectrum is sufficiently well behaved, for example when eigenvalue gaps are rationally independent [2308.02371].

## 6. Quantum-information role and analytic analogues

The physical motivation for fractional revival is consistent across the literature: it is a quantum transport phenomenon used for entanglement generation in quantum spin networks. Exact fractional revival produces a state localized on two sites, and balanced revival yields a maximally entangled pair. PGFR is weaker, but it preserves the same two-site or few-site transport architecture asymptotically [1801.09654], [2208.05107].

Relative to pretty good state transfer, PGFR is explicitly broader. The path and cycle classifications show families where PGFR occurs but PGST does not; the cycle result \(n=2p^k\) already exceeds the known PGST families, and the asymmetric path family \(n=5\cdot 2^k-1\) is presented as a new infinite family of graphs with PGFR [2005.00492]. In group-based networks, unitary Cayley graphs with \(n=2p\) and quadratic unitary Cayley graphs with \(n=8\) or \(n=2p\) provide further examples where pretty good fractional phenomena persist beyond exact or transfer-only behavior [2508.18068].

Analytic spin-chain and classical chain models clarify the boundary between exact and pretty good behavior. In \(XX\) quantum spin chains, para-Krawtchouk bi-lattice constructions and isospectral deformations realize exact fractional revival with controllable amplitudes and phases; when the bi-lattice parameter is irrational, exact revival does not occur, but the discussion states that “almost perfect” or pretty good fractional revival is possible [1507.05919]. In analytic mass-spring chains built from para-Racah polynomials, exact endpoint fractional revival is achieved by spectral commensurability, while “pretty good” revival is identified with the non-commensurate case in which revival is only approximate and occurs after very long times [2108.09386].

Taken together, these results place PGFR at the intersection of spectral graph theory, Diophantine approximation, and engineered quantum transport. The exact theory isolates the required algebraic structure; the pretty good theory shows how much of that structure can be relaxed while retaining asymptotically precise localization on designated vertices or subsets.

Source: https://www.emergentmind.com/topics/pretty-good-fractional-revival