2-Diregular Digraphs: Structure & Extremal Theory

Updated 7 July 2026

2-diregular digraphs are finite loopless (or multigraph) digraphs where every vertex has both an indegree and outdegree of 2, serving as a fundamental rigidity case in graph theory.
Near-Moore and excess theories for these digraphs provide precise bounds and nonexistence results, highlighting the delicate balance in cycle decompositions and route obstructions.
Algorithmic and algebraic investigations reveal NP-hard packing challenges alongside explicit classification via Hamiltonicity criteria, spectral circle methods, and association schemes.

A 2-diregular digraph is a digraph in which every vertex has indegree $2$ and outdegree $2$. Much of the literature uses “2-regular digraph” for the same notion, while some recent work abbreviates it to “2-dd.” As a low-degree Eulerian class, 2-diregular digraphs sit at the intersection of Hamiltonicity, branchings and arc-disjoint packing, degree/diameter extremal theory, association-scheme methods, planar embedding theory, and exact enumeration; in several of these areas, the degree-$2$ case is the first nontrivial boundary case and often the most rigid (Tuite, 2017, Messegué et al., 2024, Ramanath, 28 Jul 2025).

1. Definitions and ambient settings

In the simple-digraph setting, a 2-diregular digraph is a finite loopless digraph $G$ with $d^+(v)=d^-(v)=2$ for every vertex $v$ . In the broader enumerative setting, the same phrase may refer to a directed multigraph in which loops and multiarcs are allowed, and disconnected graphs are also counted; there the adjacency matrix $A=(a_{rc})$ is an $n\times n$ matrix with nonnegative integer entries such that every row sum and every column sum is $2$ (Mathar, 2019). The coexistence of these two models is standard in the subject: structural and extremal papers usually work with simple digraphs, while exact counting and mathematical-physics applications often admit loops and multiplicities.

Recent Hamiltonicity work uses a slightly wider notion of $k$ -digraph, allowing entry, exit, and saturated vertices, and then defines a $2$0-diregular digraph as the special case in which every vertex is saturated. In that language, a 2-diregular digraph is exactly a 2-digraph in which every vertex has $2$1 (Ramanath, 28 Jul 2025). A useful general fact is that if a $2$2-regular digraph is connected as an undirected graph, then it is strongly connected; for 2-diregular digraphs, weak connectivity and strong connectivity therefore coincide (Ramanath, 28 Jul 2025).

This terminological flexibility matters because many results for 2-diregular digraphs are model-sensitive. Planar embedding theorems are formulated for 2-cell embeddings of the underlying graph in a surface, degree/diameter results are stated for $2$3-geodetic simple digraphs, and enumerative results may include disconnected, looped, or multiply edged objects. The common invariant across these settings is the rigid local balance $2$4.

2. Near-Moore and excess theory

The degree/diameter theory of 2-diregular digraphs is organized around the directed Moore bound

$2$5

For $2$6,

$2$7

An almost Moore digraph is a diregular digraph of degree $2$8, diameter $2$9, and order $2$0. In degree $2$1, the almost Moore order is

$2$2

and the known picture is sharp: $2$3-digraphs exist for $2$4, but for $2$5 there are no almost Moore digraphs of degree $2$6 (Messegué et al., 2024).

A complementary “excess” theory studies $2$7-geodetic digraphs of order $2$8. For degree $2$9, excess $G$ 0, $G$ 1, and $G$ 2 have all been analyzed in detail. The principal facts are summarized below.

Regime	Degree-2 status	Source
Almost Moore $G$ 3	Exists for $G$ 4; no $G$ 5-digraphs for $G$ 6	(Messegué et al., 2024)
Excess $G$ 7	No $G$ 8-digraphs for $G$ 9	(Tuite, 2017, Tuite, 2017)
Excess $d^+(v)=d^-(v)=2$ 0	Any $d^+(v)=d^-(v)=2$ 1-digraph with $d^+(v)=d^-(v)=2$ 2 is diregular; there are exactly two diregular $d^+(v)=d^-(v)=2$ 3-digraphs; none exist for $d^+(v)=d^-(v)=2$ 4	(Tuite, 2017, Tuite, 2017)
Excess $d^+(v)=d^-(v)=2$ 5	Diregular $d^+(v)=d^-(v)=2$ 6-digraphs exist; there are no diregular $d^+(v)=d^-(v)=2$ 7-digraphs for $d^+(v)=d^-(v)=2$ 8	(Tuite, 2017)

These results show that near the Moore bound, degree $d^+(v)=d^-(v)=2$ 9 is exceptionally rigid. The excess- $v$ 0 theorem first forces in-regularity from the assumptions “minimum out-degree $v$ 1, $v$ 2-geodetic, order $v$ 3,” and then the excess- $v$ 4 classification collapses to exactly two isomorphism types at $v$ 5 and none beyond (Tuite, 2017, Tuite, 2017). The excess- $v$ 6 theorem extends this nonexistence frontier one step further for diregular digraphs (Tuite, 2017). This suggests that the smallest nontrivial 2-diregular $v$ 7-geodetic examples are highly exceptional objects rather than representatives of a broad parametric family.

The almost Moore literature adds a separate structural layer through repeats and self-repeats. In an almost Moore digraph, the repeat map $v$ 8 is a permutation and indeed an automorphism; if self-repeats occur and $v$ 9, there are exactly $A=(a_{rc})$ 0 self-repeats forming a directed cycle $A=(a_{rc})$ 1. Although the newest nonexistence results on self-repeats are stated for $A=(a_{rc})$ 2, the paper explicitly places $A=(a_{rc})$ 3 in the same framework and recalls that the degree- $A=(a_{rc})$ 4 case is already settled: no almost Moore digraphs exist for $A=(a_{rc})$ 5 (Messegué et al., 2024).

3. Hamiltonicity and non-Hamiltonian structure

Hamiltonicity is the most developed structural theme for 2-diregular digraphs. A useful cautionary point comes from the surrounding low-degree diregular literature: Jackson’s conjecture, as quoted in the 2016 note, explicitly excludes $A=(a_{rc})$ 6, and the counterexample exhibited there is 3-diregular rather than 2-diregular. The note therefore does not state a Hamiltonicity theorem for 2-diregular oriented graphs, and its main implication for degree $A=(a_{rc})$ 7 is negative and indirect: even a highly symmetric, oriented, circulant, strongly connected, low-degree diregular digraph can fail to be Hamiltonian (Guninski, 2016).

A much more specific structure theory is developed in recent work on 2-dds. Every 2-digraph admits a unique partition of its arc set into alternating cycles (ACs). If $A=(a_{rc})$ 8 has AC set $A=(a_{rc})$ 9, then a factor of $n\times n$ 0 is obtained by choosing, for each $n\times n$ 1, either all forward arcs $n\times n$ 2 or all backward arcs $n\times n$ 3. Consequently, a graph with $n\times n$ 4 has exactly $n\times n$ 5 factors. For a 2-diregular digraph, every factor is a disjoint union of directed cycles, the index $n\times n$ 6 of a factor is the number of cycles it contains, and

$n\times n$ 7

where $n\times n$ 8 is the minimum index over all factors (Ramanath, 28 Jul 2025). This converts Hamiltonicity into a problem about cycle counts across a Boolean family of factors determined by the AC decomposition.

The same framework introduces odd and even 2-dds: a 2-dd is odd if every factor has odd index, and even if every factor has even index. Routes and quotients refine this picture. For an open 2-digraph, an open factor defines a bijection from entry vertices to exit vertices; quotienting by a subset of ACs compresses the graph along such open routes, and Hamiltonicity behaves functorially under this compression. In particular, if $n\times n$ 9 is Hamiltonian, then every nonempty quotient $2$0 contains a Hamiltonian element; conversely, if some quotient is Hamiltonian, then $2$1 is Hamiltonian. In the special class $2$2 of 2-dds whose alternating cycles all have six arcs, the paper proves that an odd graph is non-Hamiltonian if and only if it contains a nonempty AC-induced subgraph that is closed (Ramanath, 28 Jul 2025).

A further refinement, developed for a subclass built by splicing open 2-graphs, translates non-Hamiltonicity into a permutation problem. Open route sets become uniform subsets $2$3, biconjugation is defined by $2$4, and each uniform set has an excluded set $2$5 and a residue $2$6. For a spliced 2-dd

$2$7

the paper proves the exact criterion

$2$8

This gives a necessary and sufficient condition for non-Hamiltonicity in that splicing framework and yields cardinality tests such as $2$9 Hamiltonian (Ramanath, 23 Jan 2026).

These tools are constructive as well as diagnostic. Recent work identifies several families of non-Hamiltonian 2-dds, including even members of $k$ 0, products of directed cycles, and graphs obtained by splicing closed or uniquely routed open 2-digraphs to already non-Hamiltonian 2-dds (Ramanath, 28 Jul 2025). A plausible implication is that Hamiltonicity in degree $k$ 1 is governed less by generic expansion heuristics than by very explicit route and factor obstructions.

4. Branchings, packings, and algorithmic complexity

The packing side of 2-diregular digraph theory is represented by the study of branchings. An out-branching $k$ 2 rooted at $k$ 3 is a connected spanning subdigraph in which every vertex $k$ 4 has precisely one arc entering it and $k$ 5 has no arcs entering it; the dual notion is an in-branching $k$ 6 (Bang-Jensen et al., 2012). Two natural arc-disjoint packing problems are central.

First, given a 2-regular digraph $k$ 7, deciding whether it contains two arc-disjoint branchings $k$ 8 and $k$ 9 remains NP-complete, even though both indegree and outdegree are fixed at $2$00. Second, for a 2-regular digraph $2$01, deciding whether $2$02 contains an out-branching $2$03 such that $2$04 remains connected behaves differently: it is polynomial-time solvable when the root is not prescribed in advance. In the 2-regular setting, the paper proves that this second problem is equivalent to deciding whether $2$05 contains two arc-disjoint out-branchings. The same paper also generalizes the root-free phenomenon to $2$06-regular digraphs, where one seeks $2$07 pairwise arc-disjoint spanning trees and out-branchings in total (Bang-Jensen et al., 2012).

These results place 2-diregular digraphs on a sharp algorithmic boundary. Regularity at degree $2$08 does not eliminate NP-completeness for natural directed packing problems, but it does create root-free polynomial cases and exact equivalences that do not hold in arbitrary digraphs. In that sense, 2-DDS are sparse enough to be rigid and rich enough to retain hard packing structure.

5. Algebraic and metric regularity frameworks

One major algebraic framework is that of normally regular digraphs (NRDs), defined by the matrix identity

$2$09

The paper proves that every NRD has normal adjacency matrix, and that a connected $2$10-regular digraph with normal adjacency matrix is an NRD if and only if all eigenvalues other than $2$11 lie on one circle in the complex plane (Jørgensen, 2014). Specializing the general parameter relation

$2$12

to $2$13 gives

$2$14

The same paper does not provide explicit 2-regular examples or a classification for $2$15. This suggests that 2-diregular NRDs, if they exist beyond very small or degenerate cases, are strongly constrained by arithmetic and spectral conditions (Jørgensen, 2014).

A second framework is weak distance-regularity. For diameter $2$16, every weakly distance-regular digraph has an attached association scheme with at most four nontrivial classes, and the scheme is commutative. The characterization proved in 2025 says that such a digraph has exactly one pair of non-symmetric relations $2$17, with the arc set equal either to $2$18 alone or to $2$19 in a 2-, 3-, or 4-class scheme subject to explicit exclusions of certain $2$20-polynomial, wreath-product, or wedge-product cases. The 2-diregular subcases are precisely those for which the relevant valencies satisfy $2$21 or $2$22 (Wang et al., 15 Jul 2025).

A more concrete classification is available for commutative weakly distance-regular digraphs whose underlying graphs are Hamming graphs, folded $2$23-cubes, and Doob graphs. Among the resulting examples, the only genuine 2-diregular case in the classified families is

$2$24

whose underlying graph is the 3-cube $2$25. The same paper proves that no commutative weakly distance-regular digraph has a folded $2$26-cube as its underlying graph when $2$27 (Yang et al., 2023). The valency-3 companion classification with two arc types consists entirely of explicit Cayley families, which reinforces the algebraic rigidity already visible at valency $2$28 (Yang et al., 2015).

A third framework comes from common-neighbor regularity. A two-way $2$29-liking digraph is a digraph in which every $2$30-set of vertices has exactly $2$31 common out-neighbors and exactly $2$32 common in-neighbors. For $2$33 and $2$34, every such digraph is $2$35-diregular and satisfies

$2$36

Specializing to $2$37 forces $2$38, which leaves no loopless 2-diregular example for $2$39. Together with the $2$40-liking classification, the paper concludes that the only 2-diregular two-way $2$41-liking digraph is the complete digraph $2$42, occurring at $2$43, $2$44 (Chu et al., 2024). This sharply limits how much homogeneous common-neighbor symmetry can coexist with degree $2$45.

6. Planar embeddings, enumeration, and applications

For planar and surface theory, the relevant notion of embedding is stronger than an embedding of the underlying graph. A 2-regular digraph $2$46 is embedded in a surface $2$47 by a 2-cell embedding of the underlying graph with the additional property that every face is bounded by a directed walk. Equivalently, at each vertex the four incident edges appear in the local rotation in the alternating pattern in–out–in–out (Archdeacon et al., 2017). In this setting, there is a directed analogue of Whitney’s theorem: if $2$48 and $2$49 are embeddings of a connected 2-regular digraph in $2$50, then $2$51 can be transformed into an embedding equivalent to $2$52 by a sequence of Whitney flips. There is also a directed analogue of Tutte’s peripheral-cycle theorem: every edge in a strongly 2-edge-connected Eulerian digraph is contained in at least two peripheral cycles. As a consequence, any two embeddings of a strongly 2-edge-connected 2-regular digraph in $2$53 are equivalent (Archdeacon et al., 2017).

Exact enumeration is available in the multigraph-with-loops model. The number $2$54 of unlabeled 2-regular digraphs on $2$55 vertices, including disconnected graphs and allowing loops and multiarcs, is

$2$56

These objects are represented by $2$57 nonnegative integer matrices with all row sums and all column sums equal to $2$58, and the paper further refines the counts by number of weak components, by rooted versions, and by “fairly regular” variants obtained by cutting one or two arcs (Mathar, 2019). The same paper gives a lossless transformation between Bogdanos’ bipartite contraction graphs for products of Riemann tensors and 2-regular digraphs: each unlabeled 2-regular digraph on $2$59 vertices corresponds to one contraction pattern of a product of $2$60 Riemann tensors in a Lovelock-type Lagrangian (Mathar, 2019).

Taken together, these results show that 2-diregular digraphs are simultaneously sparse, rigid, and unexpectedly versatile. In extremal graph theory they mark the failure of near-Moore optimism beyond very small parameters; in Hamiltonicity they admit decomposition theories strong enough to yield exact route-based obstructions; in algebraic graph theory they sit naturally inside association schemes and spectral circle criteria; in topological graph theory they support Whitney- and Tutte-type theorems; and in enumeration they are explicit enough to be tabulated and repurposed in mathematical physics.