Does a $\Sigma_5$-flow imply an $H_5$-flow? Unit-sphere flows versus oriented 5-cycle double covers

The source paper (arXiv:2510.19411) proves that a graph admits an $H_5$-flow — a nowhere-zero flow whose edge values are the twenty integer vectors of $\mathbb{R}^5$ with one coordinate $1$, one coordinate $-1$ and the rest $0$ — if and only if it admits an oriented 5-cycle double cover, and it asks whether the corresponding unit-sphere relaxation always suffices: a $\Sigma_5$-flow is the same as a unit-vector flow in $\mathbb{R}^4$, i.e. $\phi_4(G)=2$. In dimensions 3 and 4 the relaxation is strictly weaker — Thomassen constructed non-cubic graphs with an $S^1$-flow but no 3-flow, and the Petersen graph has a $\Sigma_4$-flow but no $H_4$-flow — and the paper asks whether dimension 5 behaves the same: does there exist a graph with a $\Sigma_5$-flow but no $H_5$-flow? By the paper's equivalence, such a graph would refute the Oriented 5-Cycle Double Cover Conjecture of Archdeacon and Jaeger, which is recorded as still open in the July 2026 expositions of OpenAI's proof of the (unoriented) Cycle Double Cover Conjecture. Conversely, a proof that $\Sigma_5$-flows always imply $H_5$-flows would reduce the Oriented 5-Cycle Double Cover Conjecture exactly to the statement that every bridgeless graph has $\phi_4(G)=2$ (the $d=4$ case of Jain's conjecture). The question is open.

Background

A nowhere-zero flow on a graph $G$ is a choice of direction for each edge together with an assignment of a nonzero value to each edge such that, at every vertex, total inflow equals total outflow. Such a flow can exist only on a bridgeless graph, meaning a graph with no bridge (an edge whose deletion disconnects the graph). Tutte's famous 5-flow conjecture predicts that every bridgeless graph admits a nowhere-zero flow with integer values of absolute value at most 4, while Seymour's 6-flow theorem guarantees that values of absolute value at most 5 always suffice. A closely related structural notion is a cycle double cover: a collection of cycles of $G$ — here a cycle is a subgraph in which every vertex has even degree, equivalently an edge-disjoint union of closed circuits — such that every edge belongs to exactly two members of the collection. The classical Cycle Double Cover Conjecture asked whether every bridgeless graph has one; in July 2026 OpenAI announced a proof, since rewritten as expositions by Geelen and by Oum. A stronger and still-open variant is the Oriented 5-Cycle Double Cover (Oriented 5-CDC) Conjecture of Archdeacon and Jaeger: every bridgeless graph admits an oriented 5-cycle double cover, that is, at most five directed cycles (Eulerian subgraphs equipped with an Eulerian orientation, in which each vertex of the subgraph has equal indegree and outdegree) such that each edge is traversed exactly once in each direction. Oriented cycle double covers matter for flows because a classical lemma (see Oum's exposition) shows that an orientable $k$-cycle double cover yields a nowhere-zero $\mathbb{Z}_k$-flow; the Oriented 5-CDC conjecture therefore implies Tutte's 5-flow conjecture.

Between the discrete and structural worlds lie vector-valued flows, studied in this form by Thomassen and by Wang, Cheng, Luo, and Zhang. Fix a real $r\ge2$ and an integer $d\ge1$: an $(r,d)$-nowhere-zero flow ($(r,d)$-NZF) on $G$ is a flow whose edge values are vectors of $\mathbb{R}^d$ with Euclidean norm between $1$ and $r-1$, and the $d$-dimensional flow number $\phi_d(G)$ is the smallest such $r$; for $d=1$ this recovers circular flows. Let $S^k$ denote the set of unit vectors of $\mathbb{R}^{k+1}$, and call a flow whose values all lie in a fixed set $X$ an $X$-flow. An $S^k$-flow is thus exactly a flow all of whose values have norm 1, i.e. a $(2,k+1)$-NZF, so $\phi_{k+1}(G)=2$ holds precisely when $G$ has an $S^k$-flow. Jain conjectured, in a 2007 Open Problem Garden post recording his conjectures, that every bridgeless graph has an $S^2$-flow — equivalently $\phi_3(G)=2$; the source paper states the general form ($\phi_d(G)=2$ for every $d\ge3$) as its Conjecture 1. These were the first of two Jain conjectures that together would imply Tutte's 5-flow conjecture; the second was refuted by Ulyanov in 2026, while the flow conjecture remains open.

The source paper, by Mattiolo, Mazzuoccolo, Rajník, and Tabarelli, gives a geometric description of $d$-dimensional flows and connects them to oriented cycle double covers. For $d\ge3$ it defines

$\Sigma_d=\Big\{(x_1,\dots,x_d)\in\mathbb{R}^d:\ \sum_{i=1}^{d}x_i=0,\ \sum_{i=1}^{d}x_i^2=2\Big\},\qquad H_d=\Sigma_d\cap\mathbb{Z}^d.$

$\Sigma_d$ is a $(d-2)$-dimensional sphere in the hyperplane $\sum x_i=0$ of $\mathbb{R}^d$, so a $\Sigma_d$-flow is the same as an $S^{d-2}$-flow; in particular a $\Sigma_5$-flow is a unit-vector flow in four dimensions, i.e. $\phi_4(G)=2$. $H_d$ consists of the $d(d-1)$ integer vectors with exactly one coordinate equal to $1$, one equal to $-1$ and the rest $0$ (so $|H_5|=20$); for $d=3$ its six points are the vertices of a regular hexagon, which is the link with Thomassen's complex flows (values among the third roots of unity). The paper's central result (its Theorem 4) is that a graph admits an $H_d$-flow if and only if it admits an oriented $d$-cycle double cover; for $d=3$ this recovers the classical equivalence between nowhere-zero 3-flows and oriented 3-cycle double covers. Two facts delimit what is known about relaxing $H_d$ to $\Sigma_d$. First, Thomassen showed that an $S^1$-flow does not force a nowhere-zero 3-flow for non-cubic graphs, so graphs exist with a $\Sigma_3$-flow but no $H_3$-flow (for cubic graphs the notions coincide). Second, the Petersen graph — the smallest bridgeless cubic graph that is not 3-edge-colourable (its edges cannot be partitioned into three perfect matchings) — admits a $\Sigma_4$-flow yet has no $H_4$-flow, since an oriented 4-cycle double cover is in particular a 4-cycle double cover, and a classical equivalence of Jaeger says a cubic graph has a 4-cycle double cover only if it is 3-edge-colourable. The source paper also proves that the Oriented 5-CDC conjecture implies $\phi_d(G)=2$ for all $d\ge4$ (its Proposition 7), and that it implies $\phi_2(G)\le\tau^2$ with $\tau$ the golden ratio (a result of the same authors' earlier paper on two-dimensional flows). Combining the paper's Theorem 4 with the sphere interpretation, the Oriented 5-CDC conjecture is exactly the conjunction of two statements: (a) every bridgeless graph has $\phi_4(G)=2$, and (b) every graph with a $\Sigma_5$-flow admits an $H_5$-flow. Statement (a) is Jain's conjecture at $d=4$; statement (b) is the open question posed here.

Problem Statement

Does there exist a finite graph $G$ that admits a $\Sigma_5$-flow but no $H_5$-flow? Equivalently: is it true that every graph $G$ with $\phi_4(G)=2$ — that is, every graph admitting an $S^3$-flow, a nowhere-zero flow by unit vectors of $\mathbb{R}^4$ — admits an oriented 5-cycle double cover? This is the first open case of the general separation problem asking for which $d$ a $\Sigma_d$-flow implies an $H_d$-flow: the implication fails for $d=3$ (Thomassen's non-cubic examples) and for $d=4$ (the Petersen graph). By Theorem 4 of the source paper, a graph with a $\Sigma_5$-flow but no $H_5$-flow would be a counterexample to the Oriented 5-Cycle Double Cover Conjecture of Archdeacon and Jaeger. A proof that every $\Sigma_5$-flow graph admits an $H_5$-flow would establish the Oriented 5-CDC conjecture for all graphs with $\phi_4(G)=2$ and, combined with the statement that every bridgeless graph has $\phi_4(G)=2$ (the $d=4$ case of Jain's conjecture), would be exactly equivalent to the full Oriented 5-CDC conjecture. Both outcomes are admissible resolutions: an explicit witness graph, or a proof of the implication.

References

1. D. Mattiolo, G. Mazzuoccolo, J. Rajník, G. Tabarelli, Geometric description of $d$-dimensional flows of a graph, arXiv:2510.19411 (2025), https://arxiv.org/abs/2510.19411
2. C. Thomassen, Group flow, complex flow, unit vector flow, and the $(2+\epsilon)$-flow conjecture, Journal of Combinatorial Theory, Series B 108 (2014), 81-91, doi:10.1016/j.jctb.2014.02.012, https://doi.org/10.1016/j.jctb.2014.02.012
3. P. D. Seymour, Nowhere-zero 6-flows, Journal of Combinatorial Theory, Series B 30 (1981), 130-135, doi:10.1016/0095-8956(81)90058-7, https://doi.org/10.1016/0095-8956(81)90058-7
4. D. Archdeacon, Face colorings of embedded graphs, Journal of Graph Theory 8 (1984), 387-398, doi:10.1002/jgt.3190080307, https://doi.org/10.1002/jgt.3190080307
5. F. Jaeger, Nowhere-zero flow problems, in: Selected Topics in Graph Theory 3, Academic Press, San Diego, 1988, pp. 71-95, MathSciNet MR1205397, https://mathscinet.ams.org/mathscinet-getitem?mr=1205397
6. D. Mattiolo, G. Mazzuoccolo, J. Rajník, G. Tabarelli, On $d$-dimensional nowhere-zero $r$-flows on a graph, European Journal of Mathematics 9 (2023), article 101, doi:10.1007/s40879-023-00694-1, https://doi.org/10.1007/s40879-023-00694-1
7. Y. Wang, J. Cheng, R. Luo, C.-Q. Zhang, Vector flows and integer flows, SIAM Journal on Discrete Mathematics 29 (2015), 2166-2178, doi:10.1137/151006329, https://doi.org/10.1137/151006329
8. K. Jain, Unit vector flows, Open Problem Garden, posted by M. DeVos on March 7, 2007, http://www.openproblemgarden.org/op/unit_vector_flows
9. N. Ulyanov, Graph Puzzles II.1: Counterexamples to Jain's Second Unit Vector Flows Conjecture, arXiv:2603.23328 (2026), https://arxiv.org/abs/2603.23328
10. C. Li, J. Li, R. Luo, B. Su, High-Dimensional $p$-Normed Flows, arXiv:2601.12036 (2026), https://arxiv.org/abs/2601.12036
11. C. Li, J. Li, R. Luo, B. Su, Reduction Operations and Structural Characterizations of $S^1$-Flows in Graphs, arXiv:2608.18725 (2026), https://arxiv.org/abs/2608.18725
12. S.-i. Oum, A proof of the cycle double cover conjecture by OpenAI: An exposition, arXiv:2607.16356 (2026), https://arxiv.org/abs/2607.16356
13. J. Geelen, OpenAI's proof of the Cycle Double Cover Theorem, arXiv:2607.15399 (2026), https://arxiv.org/abs/2607.15399
14. OpenAI, A proof of the cycle double cover conjecture (2026), https://cdn.openai.com/pdf/04d1d1e4-bc75-476a-97cf-49055cd98d31/cdc_proof.pdf
15. R. Hušek, R. Šámal, Exponentially Many Circuit Double Covers, arXiv:2607.24724 (2026), https://arxiv.org/abs/2607.24724
16. S. Kintali, Three-Bit Flows and Cycle Covers. Part I, arXiv:2607.14140 (2026), https://arxiv.org/abs/2607.14140
17. L. Gáborik, S. Kurz, G. Mazzuoccolo, J. Rajník, F. Rieg, Manhattan and Chebyshev flows, arXiv:2510.22234 (2025), https://arxiv.org/abs/2510.22234