Nash-Williams Orientation Conjecture

• Nash-Williams Orientation Conjecture is a hypothesis extending Nash-Williams’s finite theorem by asserting that every 2k-edge-connected infinite graph admits a k-arc-connected orientation.
• The conjecture investigates conditions under which finite connectivity thresholds generalize to infinite graphs, with studies addressing locally finite, one-ended, and rayless graphs.
• Recent research also explores edge-flip reconfiguration methods and the inherent algorithmic complexity, highlighting both constructive proofs and remaining open challenges.

The Nash-Williams Orientation Conjecture usually denotes the infinite extension of Nash-Williams’s 1960 finite orientation theorem: every $2k$-edge-connected infinite graph should admit a $k$-arc-connected orientation. In the finite setting, Nash-Williams proved the exact threshold $\lambda(G)\ge 2k$, while related literature also studies a stronger pathwise formulation requiring that, for every ordered pair of vertices, the directed local edge-connectivity is at least half of the undirected local edge-connectivity, rounded down (Assem et al., 2023, Pitz et al., 2024).

1. Formulation and terminology

For an undirected graph $G=(V,E)$, allowing parallel edges, the cut determined by a nonempty proper set $S\subset V$ is $\delta_G(S)$, and its size is $d_G(S)=|\delta_G(S)|$. The undirected edge-connectivity is

$\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$

Thus $G$ is $k$-edge-connected iff $k$0 for all nontrivial $k$1.

For a digraph $k$2, the directed cut parameters are

$k$3

and the directed edge-connectivity parameter is

$k$4

A digraph is $k$5-arc-strong, or $k$6-edge-connected in the directed sense, iff $k$7 and $k$8 for all nontrivial $k$9, equivalently $\lambda(G)\ge 2k$0 (Ito et al., 2021).

In this language, the finite Nash-Williams theorem states that an undirected graph admits a $\lambda(G)\ge 2k$1-arc-strong orientation iff it is $\lambda(G)\ge 2k$2-edge-connected. The phrase “Nash-Williams Orientation Conjecture” is therefore not used for the finite statement itself, which is a theorem; in contemporary infinite-graph work it refers to the conjectural extension from finite to infinite graphs (Ito et al., 2021, Assem, 2023).

A related but stronger formulation concerns local pairwise connectivity. For vertices $\lambda(G)\ge 2k$3, let $\lambda(G)\ge 2k$4 be the maximum number of pairwise edge-disjoint $\lambda(G)\ge 2k$5–$\lambda(G)\ge 2k$6 paths in $\lambda(G)\ge 2k$7, and let $\lambda(G)\ge 2k$8 be the maximum number of pairwise arc-disjoint directed $\lambda(G)\ge 2k$9 paths in an orientation $G=(V,E)$0. Nash-Williams’s strong finite theorem asserts that every finite graph has an orientation with

$G=(V,E)$1

for all distinct $G=(V,E)$2 (Pitz et al., 2024).

2. Finite theory and the exact $G=(V,E)$3 threshold

The finite orientation theorem of Nash-Williams generalizes Robbins’ theorem, the case $G=(V,E)$4, which states that a graph has a strongly connected orientation iff it is $G=(V,E)$5-edge-connected. In full generality, Nash-Williams proved:

$G=(V,E)$6

This is the precise threshold for the existence of a $G=(V,E)$7-arc-strong orientation (Ito et al., 2021).

The necessity of the $G=(V,E)$8 bound is immediate from cuts. In any $G=(V,E)$9-arc-strong orientation, every nontrivial cut must contain at least $S\subset V$0 arcs in each direction, so the underlying undirected cut has size at least $S\subset V$1. The content of Nash-Williams’s theorem is that this necessary condition is also sufficient (Assem, 2023).

The finite theory also contains the stronger “well-balanced” or “strong” orientation phenomenon. One formulation states that every finite multigraph admits an orientation $S\subset V$2 such that, for all distinct vertices $S\subset V$3,

$S\subset V$4

where $S\subset V$5 is a maximum family of pairwise edge-disjoint undirected $S\subset V$6–$S\subset V$7 paths and $S\subset V$8 is a maximum family of pairwise arc-disjoint directed $S\subset V$9 paths (Pitz et al., 2024). Another formulation, used in later infinite-graph work, expresses well-balancedness by requiring that across every finite cut the imbalance between the two directions is at most $\delta_G(S)$0 (Aurichi et al., 7 Oct 2025).

A key ingredient in Nash-Williams’s original proof is admissible odd-vertex pairings. Subsequent work showed that deciding whether a given odd-vertex pairing is admissible is co-NP-complete for two natural formalizations, even though the existence theorem itself remains valid (Hörsch, 2021).

3. The infinite conjecture and its present status

The infinite Nash-Williams Orientation Conjecture asks whether the finite $\delta_G(S)$1 threshold remains sufficient for all infinite graphs:

$\delta_G(S)$2

Here $\delta_G(S)$3 means that for every ordered pair of distinct vertices there are $\delta_G(S)$4 arc-disjoint directed paths from the first to the second (Assem, 2023).

For $\delta_G(S)$5, the existence of a strongly connected orientation of every $\delta_G(S)$6-edge-connected infinite graph goes back to Egyed (1941) (Assem et al., 2023). For general $\delta_G(S)$7, Thomassen proved in 2016 that every $\delta_G(S)$8-edge-connected infinite graph admits a $\delta_G(S)$9-arc-connected orientation (Assem, 2023). This established a universal linear bound, though not the conjectured optimal one.

The first improvement in a substantial infinite class was obtained for $d_G(S)=|\delta_G(S)|$0-ended locally finite graphs: if $d_G(S)=|\delta_G(S)|$1 is $d_G(S)=|\delta_G(S)|$2-ended and locally finite, then $d_G(S)=|\delta_G(S)|$3 suffices for a $d_G(S)=|\delta_G(S)|$4-arc-connected orientation (Assem, 2023). This bound was later sharpened to the optimal $d_G(S)=|\delta_G(S)|$5 for all locally finite graphs with countably many ends (Assem et al., 2023). The countability of the end space is used through a strengthened decomposition into boundary-linked components, followed by liftings, immersions, and Eulerian orientation extension.

A further extension removes local finiteness. Every $d_G(S)=|\delta_G(S)|$6-edge-connected graph with countably many edge-ends admits a $d_G(S)=|\delta_G(S)|$7-arc-connected orientation (Aurichi et al., 7 Oct 2025). In that work, edge-ends are equivalence classes of rays under finite edge cuts, and the proof combines bond-faithful decompositions into countable connected subgraphs with an expansion procedure that replaces infinite-degree vertices by $d_G(S)=|\delta_G(S)|$8-rays, reducing to the locally finite countably-ended case (Aurichi et al., 7 Oct 2025).

These results leave open the full weak conjecture for arbitrary infinite graphs, especially beyond countably many edge-ends. The literature explicitly treats that case as unresolved (Aurichi et al., 7 Oct 2025).

4. Monotone reorientation and flip-graph structure

A distinct development concerns constructive reorientation in finite graphs. If $d_G(S)=|\delta_G(S)|$9 is $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$0-edge-connected, then from any orientation $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$1 one can reach a $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$2-arc-strong orientation by flipping one edge at a time without ever decreasing directed edge-connectivity:

$\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$3

This gives an edge-flip-based constructive proof of Nash-Williams’s finite theorem (Ito et al., 2021).

The incremental version is sharper. If $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$4 is already $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$5-arc-strong and $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$6 is $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$7-edge-connected, then there is a monotone sequence of at most $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$8 edge flips producing an orientation $\lambda(G)=\min_{\emptyset\neq S\subset V} d_G(S).$9 with $G$0. Iterating from $G$1 up to $G$2 yields at most $G$3 flips, and in particular at most $G$4 flips from an arbitrary starting orientation. All flips can be found in polynomial time (Ito et al., 2021).

The proof is organized around submodularity of directed cuts, uncrossing, and an augmenting-path type exchange. For a fixed root $G$5, one studies the out-tight and in-tight families

$G$6

whose inclusionwise minimal members are pairwise disjoint. The potential

$G$7

vanishes exactly when the orientation is $G$8-arc-strong. Safe sources and safe sinks are extracted from minimal tight sets by min-cut computations, and flipping along a carefully chosen directed path preserves $G$9-arc-strongness while strictly decreasing $k$0 (Ito et al., 2021).

This monotone framework also yields a reconfiguration theorem. Let $k$1 be the graph whose vertices are $k$2-arc-strong orientations of $k$3, with adjacency given by a single edge flip. If $k$4 is $k$5-edge-connected, then $k$6 is connected. The diameter is $k$7. For $k$8, connectivity of $k$9 is known to hold iff $k$00 is $k$01-edge-connected; for $k$02, the $k$03-edge-connectivity theorem is the first global reachability result via single-edge flips (Ito et al., 2021).

5. Strong infinite variants, rayless graphs, and topological reformulations

Besides the weak conjecture, infinite-graph research also studies a strong Nash-Williams orientation problem: whether every infinite graph has an orientation preserving at least half of every undirected local edge-connectivity. Nash-Williams initially claimed such an extension and later retracted the claim in 1969; the general problem remains open (Pitz et al., 2024).

A major positive result holds for rayless graphs. A ray is a one-way infinite simple path, and a graph is rayless if it contains no ray as a topological subgraph. Every rayless graph admits an orientation $k$04 such that for all distinct vertices $k$05,

$k$06

Equivalently, $k$07, with the convention that infinite local connectivity must be preserved in full (Pitz et al., 2024).

The proof proceeds first for countable rayless graphs by transfinite induction on the Schmidt–Halin order of the graph. One removes a finite reducing set $k$08, orients contracted closures $k$09 of components $k$10 of $k$11, uses an auxiliary graph $k$12 to propagate directed reachability across $k$13, glues the resulting orientations through a tree-decomposition of adhesion at most $k$14, and finally lifts the orientation back from a finite contraction. A reduction via Laviolette’s bond-faithful decomposition then extends the result from countable to arbitrary rayless graphs (Pitz et al., 2024).

A different line of work reformulates the strong problem topologically. For infinite graphs, one can ask for an orientation such that the topological arc-connectivity $k$15 in $k$16 satisfies

$k$17

for all distinct vertices $k$18. This “Alternative Strong Conjecture” is proved for finitely separated graphs, and the same paper shows that if every locally finite graph has a well-balanced orientation, then every graph has one (Aurichi et al., 7 Oct 2025). These results isolate the locally finite case as the critical remaining obstacle for the strong conjecture.

6. Complexity barriers and open problems

The existence theorems coexist with substantial algorithmic hardness. For odd-vertex pairings, the decision problems CUT-ADMISSIBILITY and ORIENTATION-ADMISSIBILITY are both co-NP-complete. More generally, deciding whether a graph has an orientation satisfying arbitrary local arc-connectivity requirements is NP-complete (Hörsch, 2021). This locates a sharp barrier between existence results of Nash-Williams type and the verification of specific certificates used in classical proofs.

In finite reconfiguration, several natural problems remain unresolved. For $k$19, it is open whether $k$20-edge-connectivity suffices for the flip graph $k$21 to be connected. The optimality of the $k$22 upper bound for monotone flip sequences is also open, as is the question whether shortest monotone sequences can be found in polynomial time. A direct analogue of the $k$23 local reachability theory fails already for $k$24, and no general polynomial-time method is known for shortest paths in $k$25 when $k$26 (Ito et al., 2021).

For infinite graphs, the weak conjecture remains open beyond countably many edge-ends, and the strong conjecture remains open for general graphs with rays (Pitz et al., 2024, Aurichi et al., 7 Oct 2025). The modern structure of the subject is therefore bifurcated. On one side, the finite theory is exact, constructive, and increasingly reconfigurational. On the other, the infinite theory now has optimal $k$27 results for large classes—locally finite graphs with countably many ends, and more generally graphs with countably many edge-ends—but still lacks a full general resolution of either the weak or the strong orientation problem (Assem et al., 2023, Aurichi et al., 7 Oct 2025).