Unitary Cayley Graphs: Structure, Spectra & Revival

• Unitary Cayley graphs are Cayley graphs derived from finite rings with identity, where vertices are adjacent if their difference is a unit.
• They exhibit rich algebraic and spectral structures, with explicit constructions in cyclic cases, matrix rings, and Dedekind domain quotients.
• Recent research classifies these graphs in terms of perfectness and well-coveredness and reveals quantum phenomena like state transfer and fractional revival.

Searching arXiv for recent and foundational papers on unitary Cayley graphs to ground the article in published work. Unitary Cayley graphs are Cayley graphs defined from additive structures and unit groups. For a finite unital associative ring $R$, the unitary Cayley graph $G_R=\mathrm{Cay}(R,R^\times)$ has vertex set $R$, and two vertices $a,b\in R$ are adjacent if and only if $a-b\in R^\times$. In the cyclic case $R=\mathbb Z_n$, this gives the graph $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$, where adjacency is equivalent to $\gcd(a-b,n)=1$. Because these graphs are naturally connected to number theory, finite algebra, representation theory, and graph theory, they have remained an active research topic; recent work ranges from arithmetic and spectral formulas to complete ring-theoretic classifications of perfectness, well-coveredness, and quantum transport phenomena (Mináč et al., 2024, Defant, 2014).

1. Core definitions and standard models

For a finite ring with identity, the unitary Cayley graph is the Cayley graph of the additive group with respect to the set of units. In the ring-theoretic notation of papers, this is written either as $G_R$ or $G(R)$, with

$G_R=\mathrm{Cay}(R,R^\times)$0

For $G_R=\mathrm{Cay}(R,R^\times)$1, this specializes to the graph $G_R=\mathrm{Cay}(R,R^\times)$2, so two matrices $G_R=\mathrm{Cay}(R,R^\times)$3 are adjacent if and only if $G_R=\mathrm{Cay}(R,R^\times)$4 is invertible (Mináč et al., 2024, Chen et al., 2019).

In the classical cyclic case,

$G_R=\mathrm{Cay}(R,R^\times)$5

the graph is $G_R=\mathrm{Cay}(R,R^\times)$6-regular. When $G_R=\mathrm{Cay}(R,R^\times)$7 is prime, $G_R=\mathrm{Cay}(R,R^\times)$8 is a complete graph; when $G_R=\mathrm{Cay}(R,R^\times)$9 is a power of a prime, $R$0 is a complete $R$1-partite graph. The adjacency matrix is circulant, Hermitian, symmetric, and integral, and the graph is connected (Soni et al., 2024).

A significant commutative generalization appears for Dedekind domain quotients. If $R$2 is a Dedekind domain and $R$3 is a nonzero ideal of finite index, then the unitary Cayley graph of $R$4, denoted $R$5, is studied as a generalized totient graph. This extends the classical Euler totient Cayley graphs $R$6 and provides a framework in which clique counts, chromatic parameters, girth, and component diameters can be handled uniformly (Defant, 2014).

2. Ring-theoretic decompositions and graph structure

A recurring theme is that the graph structure is controlled by the algebraic decomposition of the underlying ring. For finite rings, the 2024 perfectness classification reduces the problem to semisimple rings via the radical and then uses the Artin–Wedderburn theorem: $R$7 where the $R$8 are finite fields and the $R$9 are matrix rings with $a,b\in R$0. The reduction relies on the fact that wreath products and direct products preserve perfectness in the required way, allowing the classification to be expressed purely in terms of the semisimplification (Mináč et al., 2024).

For finite commutative rings, decomposition into local factors likewise governs the graph. Recent work on Grover walks states that if

$a,b\in R$1

with each $a,b\in R$2 local, then for $a,b\in R$3 and $a,b\in R$4,

$a,b\in R$5

and consequently $a,b\in R$6 is isomorphic to the tensor product $a,b\in R$7 (Bhakta et al., 14 Feb 2025).

For $a,b\in R$8, the structural reduction is especially explicit. The neighborhood of a vertex depends only on its residue modulo $a,b\in R$9, the product of the distinct primes dividing $a-b\in R^\times$0, and one has

$a-b\in R^\times$1

where $a-b\in R^\times$2 is the blow-up of $a-b\in R^\times$3 of order $a-b\in R^\times$4. When $a-b\in R^\times$5 is square-free,

$a-b\in R^\times$6

and each $a-b\in R^\times$7 is a complete graph on $a-b\in R^\times$8 vertices (Cancela et al., 2012).

Specific noncommutative families also admit precise structural descriptions. For the upper triangular matrix ring $a-b\in R^\times$9, two matrices are adjacent exactly when their diagonal entries differ in every coordinate. If $R=\mathbb Z_n$0, the graph has $R=\mathbb Z_n$1 connected components, each isomorphic to $R=\mathbb Z_n$2 with $R=\mathbb Z_n$3. If $R=\mathbb Z_n$4, the graph is connected and isomorphic to the semistrong product of $R=\mathbb Z_n$5 and the antipodal graph of the Hamming graph $R=\mathbb Z_n$6, where $R=\mathbb Z_n$7 and $R=\mathbb Z_n$8 (Hołubowski et al., 2024).

3. Spectral, arithmetic, and enumerative theory

The spectrum of $R=\mathbb Z_n$9 is classical and rigid: the eigenvalues of its adjacency matrix are the Ramanujan sums,

$X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$0

This yields several consequences collected in the adjacency-algebra study: the adjacency matrix is nonsingular if and only if $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$1 is square-free, the number of nonzero eigenvalues is $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$2, the square-free part of $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$3, the nullity is $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$4, and the adjacency algebra of $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$5 is a coherent algebra. The same work shows that $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$6 is distance regular if and only if $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$7 is a prime power or $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$8 with $X_n=\mathrm{Cay}(\mathbb Z_n,\mathbb U_n)$9 an odd prime, and strongly regular if and only if $\gcd(a-b,n)=1$0 is a prime power (Reddy, 2017).

The walk structure of $\gcd(a-b,n)=1$1 has an explicit arithmetic form. For square-free $\gcd(a-b,n)=1$2, the number of $\gcd(a-b,n)=1$3-walks between vertices factors across prime divisors, and for general $\gcd(a-b,n)=1$4,

$\gcd(a-b,n)=1$5

The same paper proves that the number of ordered $\gcd(a-b,n)=1$6-tuples of units of $\gcd(a-b,n)=1$7 whose sum is congruent to $\gcd(a-b,n)=1$8 is exactly the number of walks of length $\gcd(a-b,n)=1$9 from $G_R$0 to $G_R$1 in $G_R$2, giving a direct correspondence between walk enumeration and additive number theory (Cancela et al., 2012).

For generalized totient graphs $G_R$3, clique enumeration is governed by generalized Schemmel totient functions. The number of cliques of order $G_R$4 is

$G_R$5

and in the special case $G_R$6 this becomes

$G_R$7

The same source proves that $G_R$8 is bipartite if and only if $G_R$9, where $G(R)$0 is the smallest prime ideal norm dividing $G(R)$1, and that

$G(R)$2

That paper also corrects an erroneous claim about clique domination numbers in the earlier literature and provides a counterexample to a second claim about strong domination numbers (Defant, 2014).

Spectral generalizations now extend beyond the full unit group. For a finite commutative ring $G(R)$3 and a subgroup $G(R)$4 with $G(R)$5, a $G(R)$6-unitary Cayley graph is defined by a symmetric $G(R)$7-stable generating set. Its spectrum is described by a super-Fourier transform of the superclass characteristic function, each eigenvalue has multiplicity equal to the size of the corresponding superclass, and once an indexing is fixed, the spectrum determines the graph. In special cases, the supercharacter sums specialize to Ramanujan sums, Gauss sums, and Heilbronn sums (Nguyen et al., 14 Aug 2025).

Energy-theoretic variants are also available. For any abelian group $G(R)$8 and symmetric subset $G(R)$9, the Cayley graph $G_R=\mathrm{Cay}(R,R^\times)$00 and the Cayley sum graph $G_R=\mathrm{Cay}(R,R^\times)$01 are equienergetic. Applied to finite commutative rings, this yields families $G_R=\mathrm{Cay}(R,R^\times)$02 and, in certain cases, triples $G_R=\mathrm{Cay}(R,R^\times)$03 that are integral, equienergetic, and non-isospectral (Podestá et al., 2020).

4. Perfectness, covering properties, and regularity phenomena

The most complete recent classification concerns perfectness. Let

$G_R=\mathrm{Cay}(R,R^\times)$04

with $G_R=\mathrm{Cay}(R,R^\times)$05 finite fields and $G_R=\mathrm{Cay}(R,R^\times)$06 matrix rings. Then $G_R=\mathrm{Cay}(R,R^\times)$07 is perfect if and only if one of the following holds: $G_R=\mathrm{Cay}(R,R^\times)$08, in which case $G_R=\mathrm{Cay}(R,R^\times)$09 is bipartite; $G_R=\mathrm{Cay}(R,R^\times)$10 and $G_R=\mathrm{Cay}(R,R^\times)$11, so the semisimplification is a product of at most two finite fields; or $G_R=\mathrm{Cay}(R,R^\times)$12. Every other case is not perfect. The proof uses the Strong Perfect Graph Theorem together with explicit induced odd cycles, and the paper states that this settles the question raised by Sophie Spirkl (Mináč et al., 2024).

For matrix rings, the same classification is sharply restrictive: $G_R=\mathrm{Cay}(R,R^\times)$13 is perfect if and only if $G_R=\mathrm{Cay}(R,R^\times)$14 and $G_R=\mathrm{Cay}(R,R^\times)$15. In all other cases, either $G_R=\mathrm{Cay}(R,R^\times)$16 or $G_R=\mathrm{Cay}(R,R^\times)$17, and induced $G_R=\mathrm{Cay}(R,R^\times)$18-cycles show imperfection (Mináč et al., 2024).

Well-coveredness exhibits a different but comparably rigid pattern. For a finite field $G_R=\mathrm{Cay}(R,R^\times)$19, the unitary Cayley graph of $G_R=\mathrm{Cay}(R,R^\times)$20 is well-covered if and only if $G_R=\mathrm{Cay}(R,R^\times)$21, and its independence number is

$G_R=\mathrm{Cay}(R,R^\times)$22

For a general finite ring $G_R=\mathrm{Cay}(R,R^\times)$23, $G_R=\mathrm{Cay}(R,R^\times)$24 is well-covered if and only if $G_R=\mathrm{Cay}(R,R^\times)$25 is isomorphic to one of: a finite field; a product of two finite fields; the matrix ring $G_R=\mathrm{Cay}(R,R^\times)$26 for some finite field $G_R=\mathrm{Cay}(R,R^\times)$27; or $G_R=\mathrm{Cay}(R,R^\times)$28 for some $G_R=\mathrm{Cay}(R,R^\times)$29 (Rahimi et al., 2023).

Strong regularity for matrix algebras is completely classified. The unitary Cayley graph $G_R=\mathrm{Cay}(R,R^\times)$30 is strongly regular if and only if $G_R=\mathrm{Cay}(R,R^\times)$31. If $G_R=\mathrm{Cay}(R,R^\times)$32, then the $G_R=\mathrm{Cay}(R,R^\times)$33 case has parameters

$G_R=\mathrm{Cay}(R,R^\times)$34

For $G_R=\mathrm{Cay}(R,R^\times)$35, strong regularity fails because the number of common neighbors of non-adjacent vertices depends on the rank of the matrix difference $G_R=\mathrm{Cay}(R,R^\times)$36 (Chen et al., 2019).

5. Generalizations and related families

The modern $G_R=\mathrm{Cay}(R,R^\times)$37-unitary theory broadens the classical construction from $G_R=\mathrm{Cay}(R,R^\times)$38 to arbitrary subgroups $G_R=\mathrm{Cay}(R,R^\times)$39. In the commutative case, the generating set must satisfy $G_R=\mathrm{Cay}(R,R^\times)$40; in the noncommutative case, the condition is $G_R=\mathrm{Cay}(R,R^\times)$41. This framework unifies Paley graphs, unitary Cayley graphs, $G_R=\mathrm{Cay}(R,R^\times)$42-unitary graphs, involutory graphs, cubelike graphs, and gcd-graphs. In the noncommutative extension, the relevant orbits are double cosets $G_R=\mathrm{Cay}(R,R^\times)$43, and for $G_R=\mathrm{Cay}(R,R^\times)$44 with $G_R=\mathrm{Cay}(R,R^\times)$45, the double cosets are classified by matrix rank, so the associated unitary Cayley graphs have exactly $G_R=\mathrm{Cay}(R,R^\times)$46 distinct eigenvalues (Nguyen et al., 14 Aug 2025, Nguyen et al., 22 Mar 2026).

That same noncommutative theory gives ring-theoretic criteria for connectedness and primeness. For $G_R=\mathrm{Cay}(R,R^\times)$47 with $G_R=\mathrm{Cay}(R,R^\times)$48, every nonempty $G_R=\mathrm{Cay}(R,R^\times)$49-unitary Cayley graph is connected. The paper further proves that no gcd-graph over $G_R=\mathrm{Cay}(R,R^\times)$50 with $G_R=\mathrm{Cay}(R,R^\times)$51 admits perfect state transfer; the same negative result holds for $G_R=\mathrm{Cay}(R,R^\times)$52-unitary Cayley graphs and for $G_R=\mathrm{Cay}(R,R^\times)$53 when $G_R=\mathrm{Cay}(R,R^\times)$54 is a local commutative finite Frobenius ring (Nguyen et al., 22 Mar 2026).

Several closely related graph families are studied in parallel. The unitary addition Cayley graph $G_R=\mathrm{Cay}(R,R^\times)$55 joins distinct $G_R=\mathrm{Cay}(R,R^\times)$56 when $G_R=\mathrm{Cay}(R,R^\times)$57 is a unit. For a finite commutative ring of odd cardinality decomposed as a product of local rings $G_R=\mathrm{Cay}(R,R^\times)$58 with maximal ideals $G_R=\mathrm{Cay}(R,R^\times)$59,

$G_R=\mathrm{Cay}(R,R^\times)$60

For $G_R=\mathrm{Cay}(R,R^\times)$61 with odd $G_R=\mathrm{Cay}(R,R^\times)$62 and $G_R=\mathrm{Cay}(R,R^\times)$63 distinct odd prime factors,

$G_R=\mathrm{Cay}(R,R^\times)$64

while if $G_R=\mathrm{Cay}(R,R^\times)$65 is even, $G_R=\mathrm{Cay}(R,R^\times)$66 is bipartite with clique number and chromatic number equal to $G_R=\mathrm{Cay}(R,R^\times)$67 (Calhoun et al., 2024).

Another extension is the generalized unit and unitary Cayley graph $G_R=\mathrm{Cay}(R,R^\times)$68, where adjacency is defined by the existence of $G_R=\mathrm{Cay}(R,R^\times)$69 such that $G_R=\mathrm{Cay}(R,R^\times)$70. In this framework, the unitary Cayley graph $G_R=\mathrm{Cay}(R,R^\times)$71 is projective if and only if $G_R=\mathrm{Cay}(R,R^\times)$72 or $G_R=\mathrm{Cay}(R,R^\times)$73. The same source proves that for an Artinian ring $G_R=\mathrm{Cay}(R,R^\times)$74, finite nonorientable genus forces $G_R=\mathrm{Cay}(R,R^\times)$75 to be finite, and for a fixed positive integer $G_R=\mathrm{Cay}(R,R^\times)$76, only finitely many finite rings produce graphs of nonorientable genus $G_R=\mathrm{Cay}(R,R^\times)$77 (Khorsandi et al., 2020).

A semiring analogue dispenses with subtraction. For a semiring $G_R=\mathrm{Cay}(R,R^\times)$78, the unitary Cayley graph $G_R=\mathrm{Cay}(R,R^\times)$79 declares $G_R=\mathrm{Cay}(R,R^\times)$80 and $G_R=\mathrm{Cay}(R,R^\times)$81 adjacent when there exists $G_R=\mathrm{Cay}(R,R^\times)$82 such that $G_R=\mathrm{Cay}(R,R^\times)$83 or $G_R=\mathrm{Cay}(R,R^\times)$84. For matrix semirings $G_R=\mathrm{Cay}(R,R^\times)$85, the cited work gives bounds for diameter, clique number, independence number, and girth, and proves in particular that $G_R=\mathrm{Cay}(R,R^\times)$86 for $G_R=\mathrm{Cay}(R,R^\times)$87 (Dolžan, 2023).

6. Quantum walks, state transfer, and fractional revival

Quantum dynamics on unitary Cayley graphs has become a substantial subfield. For Grover walks on $G_R=\mathrm{Cay}(R,R^\times)$88, periodicity occurs if and only if

$G_R=\mathrm{Cay}(R,R^\times)$89

for non-negative integers $G_R=\mathrm{Cay}(R,R^\times)$90 with $G_R=\mathrm{Cay}(R,R^\times)$91. Within this class, perfect state transfer occurs only for the four graphs $G_R=\mathrm{Cay}(R,R^\times)$92, $G_R=\mathrm{Cay}(R,R^\times)$93, $G_R=\mathrm{Cay}(R,R^\times)$94, and $G_R=\mathrm{Cay}(R,R^\times)$95 (Bhakta et al., 2024).

For unitary Cayley graphs over finite commutative rings, the Grover-walk classification is ring-theoretic. If

$G_R=\mathrm{Cay}(R,R^\times)$96

with local factors $G_R=\mathrm{Cay}(R,R^\times)$97 and maximal ideals $G_R=\mathrm{Cay}(R,R^\times)$98, then $G_R=\mathrm{Cay}(R,R^\times)$99 is periodic if and only if either all residue fields $R$00, or $R$01 and $R$02 for $R$03. Perfect state transfer occurs if and only if

$R$04

(Bhakta et al., 14 Feb 2025).

For continuous-time dynamics generated by the adjacency matrix, the literature has evolved rapidly. A 2024 paper states that quantum fractional revival in unitary Cayley graphs exists only when the number of vertices is even and gives explicit examples at $R$05 with revival times $R$06, respectively (Soni et al., 2024). A later classification sharpens this: $R$07 admits fractional revival and pretty good fractional revival if and only if $R$08 or $R$09 with $R$10 prime, and for this family fractional revival and pretty good fractional revival coincide (Kalita et al., 25 Aug 2025).

The 2026 refinement gives a closed-form description for the case $R$11, with $R$12 an odd prime. The minimum revival time is

$R$13

and the revival amplitudes between antipodal vertices are

$R$14

The same paper proves that for regular graphs the Laplacian and adjacency Hamiltonians differ only by a global phase factor, that strongly cospectral pairs in unitary Cayley graphs are exactly antipodal pairs when $R$15 is even, and that for $R$16 the entanglement entropy at revival depends only on $R$17 and $R$18 (Abdullah, 13 May 2026).

Taken together, these developments show that unitary Cayley graphs form a rare class in which additive algebra, unit theory, product decompositions, exact spectral formulas, and quantum transport criteria can all be expressed in explicit arithmetic terms. Recent classifications of perfectness and revival phenomena indicate that the sharpest structural behavior occurs only in highly constrained semisimple and cyclic configurations, especially products of a very small number of fields and the exceptional matrix ring $R$19 (Mináč et al., 2024).