Weak 3-Flow Conjecture in Graph Theory
- Weak 3-Flow Conjecture is a central proposition in nowhere-zero flow theory, asserting that high edge-connectivity ensures the existence of a nowhere-zero 3-flow in graphs.
- Formulations vary: Jaeger’s version posits a threshold h (settled at 6-edge-connected), while Kochol’s and Tutte’s conjectures focus on 5- and 4-edge-connected graphs, highlighting nuanced equivalences.
- Recent progress leverages flow-extension techniques and structural decompositions to manage small-cut obstructions, yet the 5-edge-connected case remains unresolved.
Searching arXiv for recent and foundational papers on the Weak 3-Flow Conjecture and closely related notions. {"queries":[{"query":"\"weak 3-flow conjecture\" graph theory arXiv","preferred_count":10},{"query":"Tutte 3-flow conjecture Z3-connectivity arXiv","preferred_count":10}]} The Weak 3-Flow Conjecture belongs to nowhere-zero flow theory and asks whether sufficiently strong edge-connectivity forces the existence of a nowhere-zero $3$-flow. The term is not uniform in the literature. In Jaeger’s 1979 sense, it asks whether there exists an integer such that every -edge-connected graph admits a nowhere-zero $3$-flow. In Kochol’s formulation, it asks whether every $5$-edge-connected graph admits a nowhere-zero $3$-flow, a statement that is equivalent to Tutte’s $3$-flow conjecture that every $4$-edge-connected graph admits such a flow. Closely related formulations use modulo-$3$ orientations and -connectivity, and much of the modern theory moves between these equivalent or stronger perspectives (Chen et al., 2014, Li, 2020).
1. Definitions and basic equivalences
Let 0 be a finite loopless graph, possibly with multiple edges, and let 1 be an orientation of 2. A nowhere-zero 3-flow on 4 is a mapping
5
such that for every vertex 6,
7
Equivalently, one may use the integral formulation with values in 8 and exact conservation at each vertex. A modulo 9-orientation is an orientation 0 satisfying
1
for all vertices 2. By Tutte’s characterization, a graph has a nowhere-zero 3-flow if and only if it has a modulo 4-orientation (Chen et al., 2014).
A graph is 5-edge-connected if every edge-cut has size at least 6, and it is bridgeless if it has no 7-edge-cut. For a vertex set 8, the associated cut is
9
A $3$0-edge-cut is a cut of size $3$1 (Chen et al., 2014).
The stronger notion of $3$2-connectivity is defined through arbitrary boundary functions. If $3$3 satisfies
$3$4
then $3$5 is $3$6-connected if there exist an orientation $3$7 and a function
$3$8
such that for every vertex $3$9,
$5$0
Taking $5$1 yields a modulo $5$2-orientation, so $5$3-connectivity implies the existence of a nowhere-zero $5$4-flow (Chen et al., 2014).
2. Formulations and logical structure
The central difficulty of the subject is not the existence of some connectivity threshold, but the determination of the sharp threshold and the precise role of group-connectivity. The literature uses “Weak 3-Flow Conjecture” in more than one sense. In the standard Jaeger usage, it is the existence of a fixed threshold $5$5. In Kochol’s usage, or in papers that identify the weak problem with Tutte’s conjecture, the phrase may instead refer to the $5$6-edge-connected or even $5$7-edge-connected statement. This terminological instability is explicit in the literature: some later papers refer to Tutte’s $5$8-flow conjecture itself as the weak $5$9-flow conjecture (Chen et al., 2014, Mkrtchyan, 2020, Yu, 1 Oct 2025).
| Formulation | Meaning | Relation/status |
|---|---|---|
| Tutte’s $3$0-flow conjecture | Every $3$1-edge-connected graph has a nowhere-zero $3$2-flow | Open |
| Jaeger’s weak form | There exists $3$3 such that every $3$4-edge-connected graph has a nowhere-zero $3$5-flow | Settled with $3$6 |
| Kochol’s $3$7-edge-connected form | Every $3$8-edge-connected graph has a nowhere-zero $3$9-flow | Equivalent to Tutte’s conjecture |
| JLPT conjecture | Every $3$0-edge-connected graph is $3$1-connected | Implies Kochol’s form |
Thomassen proved that every $3$2-edge-connected graph is $3$3-connected, and Lovász, Thomassen, Wu, and Zhang improved this to $3$4-edge-connected graphs. Consequently, Jaeger’s existence-of-$3$5 conjecture is resolved with $3$6, but this does not settle either the JLPT $3$7-edge-connected $3$8-connectivity conjecture or Kochol’s equivalent $3$9-edge-connected version of Tutte’s conjecture (Chen et al., 2014, Hasanvand, 2016).
Kochol also proved that Tutte’s $4$0-flow conjecture is equivalent to the statement that every bridgeless graph with at most three $4$1-edge-cuts admits a nowhere-zero $4$2-flow. This recasts the conjecture as a statement about controlling the smallest odd cuts rather than simply increasing global connectivity (Chen et al., 2014).
3. Progress near the $4$3-edge-connected frontier
A major line of progress has been to verify the conjectural conclusions under restrictions on the number of small cuts. Chen and Ning proved that every $4$4-edge-connected graph with at most seven $4$5-edge-cuts admits a nowhere-zero $4$6-flow, that every bridgeless graph containing no $4$7-edge-cuts but at most three $4$8-edge-cuts admits a nowhere-zero $4$9-flow, and that every $3$0-edge-connected graph with at most five $3$1-edge-cuts is $3$2-connected. Their unifying structural statement is that if $3$3 is the set of $3$4-edge-cuts and $3$5 the set of $3$6-edge-cuts, then
$3$7
implies the existence of a modulo $3$8-orientation. The bound “at most three $3$9-edge-cuts” is sharp in the second theorem because 0, and more generally graphs contractible to 1, provide the obstruction at four 2-cuts (Chen et al., 2014).
The proof strategy in this small-cut regime is technically centered on pre-orientation extension. The decisive tool is the Lovász–Thomassen–Wu–Zhang extension lemma, which starts with a prescribed orientation at one distinguished vertex and extends it to a global 3-boundary orientation under cut inequalities of the form
4
Chen and Ning combine minimal counterexamples, contractions of small-cut sides, and an auxiliary-vertex augmentation that controls the number of degree-5 and degree-6 vertices. Their analysis makes explicit why bounding the number of 7-edge-cuts is decisive: too many 8-cuts obstruct the universal cut inequalities required by the extension lemma (Chen et al., 2014).
A different structural sufficient condition is the existence of four edge-disjoint spanning trees. Han, Lai, and Li proved that every graph with at least four edge-disjoint spanning trees is 9-connected. They also proved that every 00-edge-connected essentially 01-edge-connected graph is 02-extendable at any degree-five vertex. Their paper isolates a sharper conjectural threshold: if every 03-edge-connected essentially 04-edge-connected graph were 05-extendable at a degree-five vertex, then Jaeger–Linial–Payan–Tarsi would follow (Han et al., 2016).
Hasanvand showed that, conditional on the JLPT conjecture, every 06-tree-connected graph is 07-connected. Unconditionally, the same circle of ideas already gives that every 08-tree-connected graph is 09-connected, and the conditional result extends the conjectural scope from edge-connectivity to spanning-tree packings. The threshold is sharp in the sense that 10 is 11-tree-connected but has no nowhere-zero 12-flow (Hasanvand, 2016).
Flow-extension methods also yield near-planar progress. Li proved that the 13-flow conjecture is equivalent to the statement that every 14-edge-connected graph is 15-extendable at every 16-vertex, and that the 17-group-connectivity conjecture is equivalent to 18-extendability at every 19-vertex. The same paper verified 20-connectivity for 21-edge-connected graphs with crossing number at most one, and provided small-cut results including that every 22-edge-connected graph with at most five 23-cuts and no 24-cuts is 25-connected, and that every 26-edge-connected graph with at most seven 27-cuts is 28-connected (Li, 2020).
4. Verified subclasses beyond generic connectivity
Symmetry provides one route below the general 29-edge threshold. Li and Zhou proved that every regular graph of valency at least four admitting a solvable arc-transitive group of automorphisms admits a nowhere-zero 30-flow. In the vertex-transitive setting, Watkins’ theorem implies that a graph of valency 31 is 32-edge-connected, so the generic 33-edge-connected theorem already covers valencies at least 34, and even valency 35 is handled by nowhere-zero 36-flows. The valency-37 case is therefore the pivotal open instance within highly symmetric graphs, and the solvable arc-transitive theorem settles that case for a broad family by inductive use of normal quotients and multicover lifting (Li et al., 2013).
Another route is to constrain the independence number. A 2017 paper characterized graphs with independence number at most 38 that admit a nowhere-zero 39-flow, using two finite exceptional families 40 and 41. As a consequence, every 42-edge-connected graph of order at least 43 with 44 admits a nowhere-zero 45-flow. The same work proved that every odd-46-edge-connected graph with 47 admits a nowhere-zero 48-flow, thereby strengthening the 49-edge-connected statement in that class. A new reduction method for odd wheels, based on controlled wheel contractions that preserve the relevant connectivity, is central to the proofs (Li et al., 2017).
Cayley graphs form a third major verified family. In 2026, Ahanjideh and Kovács proved that every connected Cayley graph of valency at least 50 on a solvable group of order 51, where 52 is square-free, admits a nowhere-zero 53-flow. The difficult case is valency 54, where they introduced a necessary and sufficient criterion for a simple 55-valent graph to admit a nowhere-zero 56-flow: the vertex set must admit a partition into nonempty subsets 57 and 58 such that the induced subgraphs 59 and 60 are pseudoforests and either all components are unicyclic or suitable transversals in 61 and 62 are joined by a perfect matching. This yields a complete structural criterion for the 63-regular case within that framework (Ahanjideh et al., 25 Mar 2026).
5. Reformulations, approximations, and alternative proof paradigms
One influential reformulation replaces nowhere-zero flows by flows with a small null set. For an abelian-group flow 64, the nullity is
65
Esperet, de Verclos, Le, and Thomassé proved that Tutte’s 66-flow conjecture is equivalent to the existence of a sublinear function 67 such that every 68-edge-connected graph admits a 69-flow 70 with
71
Thus an approximate result with 72 zero-edges is exactly as strong as a true nowhere-zero theorem. The same paper gave analogous equivalences for Tutte’s 73- and 74-flow conjectures, placing the weak 75-flow problem inside a broader approximation program for Tutte-type flow conjectures (Mkrtchyan, 2020).
A second reformulation is local extension at a single vertex. Li showed that Tutte’s 76-flow conjecture is equivalent to the statement that every 77-edge-connected graph is 78-extendable at every 79-vertex, and that the 80-group-connectivity conjecture is equivalent to 81-extendability at every 82-vertex. These equivalences use a six-vertex gadget 83 together with deficiency counting and Mader-type splitting lemmas. They recast a global existence problem as a constrained pre-orientation extension problem, aligning the conjecture with the extension machinery developed by Thomassen and by Lovász–Thomassen–Wu–Zhang (Li, 2020).
A third reformulation is algebraic. A 2025 paper on the Additive Basis Conjecture proved that in a vector space over 84, the union of any four linear bases is an additive basis, establishing the 85 case with 86. The paper explicitly presents this as an alternative algebraic proof of what it calls the weak 87-flow conjecture, namely the statement that every 88-edge-connected graph has a nowhere-zero 89-flow. Its route passes through the Jaeger–Linial–Payan–Tarsi additive-basis scheme and the cycle space over 90 (Yu, 1 Oct 2025).
Taken together, these reformulations show that the weak 91-flow problem is simultaneously a cut-structure problem, an extension problem, an approximation problem, and an algebraic basis problem. This suggests that the conjecture is less a single isolated assertion than a convergence point for several different theories of graph constraints and group-valued circulations.
6. Critical obstructions and current status
Minimal obstructions are encoded by 92-flow-critical graphs. A bridgeless graph 93 is 94-flow-critical if it has no nowhere-zero 95-flow but every contraction 96 does. Kochol’s equivalence implies that Tutte’s 97-flow conjecture is equivalent to the assertion that every 98-flow-critical graph contains a vertex of degree three. Li, Luo, Ma, and Zhang proved several structural constraints on such graphs: they are 99-edge-connected, essentially 00-edge-connected, and 01-reduced, and the subgraph induced by degree-02 vertices contains no cycle unless the graph is an odd wheel. They also established density bounds
03
for an 04-vertex 05-flow-critical graph, with equality in either direction if and only if 06, and conjectured the sharper upper bound 07 for 08 (Li et al., 2020).
The lower-density side was sharpened in 2024. A paper on the sparsity of 09-flow-critical graphs proved that every 10-vertex 11-flow-critical graph other than 12 and 13 has at least
14
edges, and that this is tight up to lower-order terms. The proof uses a linear-algebraic independence argument over 15 and generalizes Koester’s planar result on the maximum average degree of 16-critical planar graphs. The same work derives an algorithmic corollary: for a connected graph with at most 17 edges, one can in time 18 either find a nowhere-zero 19-flow, certify nonexistence, or identify an edge whose contraction preserves flow existence (Dvořák et al., 2024).
The present status is therefore sharply bifurcated. The existence-of-a-threshold version of the Weak 3-Flow Conjecture is settled: every 20-edge-connected graph is 21-connected and hence has a nowhere-zero 22-flow. By contrast, the full 23-edge-connected case remains open, both in the stronger JLPT form “every 24-edge-connected graph is 25-connected” and in Kochol’s equivalent 26-edge-connected nowhere-zero-27-flow form, and therefore Tutte’s 28-edge-connected conjecture remains open as well. The verified subclasses and partial theorems strongly support the view that small cuts, especially of sizes 29, 30, and 31, are the principal obstructions, but the general transition from 32- to 33-edge-connectivity is still unresolved (Chen et al., 2014).