---
title: Normal Edge-Coloring in Cubic Graphs
url: https://www.emergentmind.com/topics/normal-edge-coloring
type: topic
---

# Normal Edge-Coloring in Cubic Graphs

Searching arXiv for recent and foundational papers on normal edge-coloring of cubic graphs.
Normal edge-coloring is a refinement of proper edge-coloring for cubic graphs in which the local color pattern around every edge is restricted to two extremal types. If \(G\) is cubic and \(\varphi\) is a proper edge-coloring, then for an edge \(e=uv\) one studies the colors seen at the ends of \(e\): equivalently, either the set consisting of \(e\) and its four adjacent edges has size \(3\) or \(5\), or the set on the four neighboring edges alone has size \(2\) or \(4\). An edge is called poor in the former case and rich in the latter, and a coloring is normal if every edge is poor or rich. The associated invariant, the normal chromatic index \(\chi'_N(G)\), asks for the minimum number of colors in such a coloring. In bridgeless cubic graphs, the statement \(\chi'_N(G)\le 5\) is equivalent to Jaeger’s Petersen Coloring Conjecture, which places normal edge-coloring at the intersection of edge-coloring, nowhere-zero flows, and the conjectural framework around Petersen colorings, cycle double covers, and Berge–Fulkerson coverings [1804.09449].

## 1. Local definitions and equivalent formulations

Let \(G\) be a cubic graph and let \(\varphi:E(G)\to [k]\) be a proper edge-coloring. For a vertex \(v\), write
\[
S_\varphi(v)=\{\varphi(e): e\in E_G(v)\}.
\]
For an edge \(e=uv\), one standard formulation defines
\[
S_e=\{\varphi(e)\}\cup \{\varphi(f): f\in (E_G(u)\setminus\{e\})\cup(E_G(v)\setminus\{e\})\}.
\]
Then \(e\) is poor if \(|S_e|=3\), rich if \(|S_e|=5\), and normal if \(|S_e|\in\{3,5\}\). Since \(G\) is cubic, the only excluded case is \(|S_e|=4\), often called abnormal [1804.09449].

A second formulation, common in more recent work, excludes the color of \(e\) itself and records only the colors on the four neighboring edges. In that notation, a normal coloring satisfies \(|S_e|\in\{2,4\}\) for every edge. The two formulations are equivalent: they differ only by whether the color of \(e\) is included in the local set [2410.01049].

The invariant
\[
\chi'_N(G)=\min\{k:\text{$G$ admits a normal $k$-edge-coloring}\}
\]
is the normal chromatic index. Two extremal examples anchor the theory. Any proper \(3\)-edge-coloring of a cubic graph is normal, with all edges poor. Any strong edge-coloring is also normal, with all edges rich. Thus normal edge-coloring interpolates between ordinary edge-coloring and strong edge-coloring rather than coinciding with either of them [1804.09449].

This local viewpoint is specific to cubic structure. Because every vertex has degree \(3\), each edge has exactly four adjacent edges, and the dichotomy between poor and rich is rigid. A plausible implication is that many proofs in the area are naturally local-to-global: they build global colorings by forcing compatible poor/rich patterns across cuts, matchings, cores, or gadgets.

## 2. Petersen coloring, normal \(5\)-colorings, and conjectural significance

Jaeger’s notion of \(H\)-coloring maps edges of a cubic graph \(G\) to edges of another cubic graph \(H\) in a way that preserves the local incidence pattern at each vertex. For the Petersen graph \(P_{10}\), Jaeger proved the equivalence
\[
P_{10}<G \quad\Longleftrightarrow\quad G \text{ admits a normal }5\text{-edge-coloring},
\]
so the Petersen Coloring Conjecture is exactly the assertion that every bridgeless cubic graph has a normal \(5\)-edge-coloring [1911.06759].

This equivalence gives normal edge-coloring its central role. In the bridgeless cubic setting, proving \(\chi'_N(G)\le 5\) would settle the Petersen Coloring Conjecture and, as recalled in the literature, would imply the Cycle Double Cover Conjecture and the Berge–Fulkerson Conjecture [1804.09449].

The relation to Berge–Fulkerson is especially concrete. If \(G\) has a normal \(5\)-edge-coloring, then \(G\) has a Petersen coloring; pulling back the six perfect matchings of the Petersen graph yields six perfect matchings in \(G\) such that every edge of \(G\) lies in exactly two of them. In that sense, a normal \(5\)-edge-coloring is not merely a coloring statement but a certificate of highly structured perfect-matching coverage [1904.02661].

A common misconception is that normal edge-coloring is only interesting for snarks. In fact the notion is defined for all cubic graphs, and \(3\)-edge-colorable graphs satisfy it trivially. The reason the literature concentrates on snarks is sharper: every \(3\)-edge-colorable cubic graph already has \(\chi'_N(G)\le 3\), so the conjectural barrier begins precisely where ordinary Tait coloring fails.

## 3. General bounds and the flow-based framework

The broadest unconditional theorem is that every simple cubic graph admits a normal \(7\)-edge-coloring, and this bound is best possible [1804.09449]. The proof is flow-theoretic. For bridgeless cubic graphs, Jaeger’s \(8\)-flow theorem gives a nowhere-zero \(\mathbb{Z}_2^3\)-flow; using the seven nonzero group elements as colors yields a normal \(7\)-edge-coloring because, at each vertex, the third incident value is determined by the other two, which forces the local poor/rich dichotomy [1804.09449].

The same paper shows that the bound \(7\) cannot be lowered for all simple cubic graphs. The key obstruction is the graph \(K\) obtained from \(K_4\) by subdividing one edge once. If a cubic graph contains \(K\) as a subgraph, then in any normal edge-coloring all edges of \(K\) are rich and must receive pairwise distinct colors; consequently such a graph satisfies \(\chi'_N(G)\ge 7\) [1804.09449].

The flow machinery is more refined than the global \(7\)-color statement suggests. In \(4\)-edge-connected graphs, special nowhere-zero \(\mathbb{Z}_2^2\)-flows can be prescribed so that selected edges receive equal or unequal values, and these prescriptions are lifted through contractions by perfect matchings to obtain rich or poor constraints in cubic graphs. This produces local control over individual edges and pairs of adjacent edges, not merely existence of a normal coloring [1804.09449].

A second line of work narrows the gap from \(7\) to \(6\) for important bridgeless subclasses. Every claw-free bridgeless cubic graph, every permutation snark, and every treelike snark admits a normal \(6\)-edge-coloring [1903.06043]. More recently, every bridgeless cubic graph with oddness at most \(2\) was shown to admit a normal \(6\)-edge-coloring, extending the earlier cycle-permutation result to all oddness-\(2\) bridgeless cubic graphs [2508.20565].

These \(6\)-color theorems use different structural inputs. The claw-free case relies on Oum’s decomposition into triangles and strings of diamonds. The permutation-snark proof compresses a \(7\)-color flow coloring by merging two carefully separated colors. The oddness-\(2\) theorem instead starts from a \(2\)-factor with exactly two odd cycles, finds two “good” paths between them, colors the complementary perfect matching with colors \(1\) and \(2\), and colors the cycles alternately with \(3,4\) and \(5,6\), checking edge by edge that every local set has size \(2\) or \(4\) [1903.06043; 2508.20565].

## 4. Quantitative normality and the role of structural parameters

A different approach asks not for full normality but for a coloring in which most edges are normal. The parameter \(\mu_3(G)\), introduced by Steffen, measures how far a cubic graph is from being covered by three perfect matchings. Given three \(1\)-factors \(M_1,M_2,M_3\), let \(E_0\) be the set of uncovered edges; then
\[
\mu_3(G)=\min |E_0|.
\]
If \(G_c\) is the corresponding \(\mu_3\)-core, the subgraph induced by \(E_0\cup E_2\) is a disjoint union of cycles, and a bound of Kaiser, Král, and Norine gives \(\mu_3(G)\le |E(G)|/5\) for every bridgeless cubic graph [1911.06759].

Using this parameter, every bridgeless cubic graph \(G\) admits a proper \(5\)-edge-coloring with at least
\[
|E(G)|-\mu_3(G)
\]
normal edges, and hence with at least \((4/5)|E(G)|\) normal edges uniformly [1911.06759]. The proof fixes a \(\mu_3\)-core, colors all edges outside the core with a “major-coloring” from \(\{1,2,3\}\), decomposes the core into even cycles, special strings assembled into waves, and remaining odd cycles, and tracks the balance of normal versus abnormal behavior with an auxiliary function
\[
\theta_{G_c,\psi}:E(G)\to\{-1,0,1\}.
\]
A key lemma shows that if \(G_c\) is a \(k\)-core and \(\psi\) is a proper edge-coloring, then the number of abnormal edges equals \(k-\theta(G)\). Thus proving \(\theta(G)\ge 0\) bounds the abnormal set by \(\mu_3(G)\) [1911.06759].

This theorem sharpens an earlier \(6\)-color approximation result: every bridgeless cubic graph admits a proper \(6\)-edge-coloring in which at least \((7/9)|E(G)|\) edges are normal [1903.06043]. The two statements are qualitatively different. The \(6\)-color bound is obtained by starting from a nowhere-zero \(\mathbb{Z}_2^3\)-flow and merging one color class; the \(5\)-color bound is obtained by analyzing an optimal triple of perfect matchings and confining abnormality to the uncovered core edges.

Several immediate consequences follow. If \(G\) is \(3\)-edge-colorable, then \(\mu_3(G)=0\), so the \(\mu_3\)-theorem recovers a normal coloring of all edges. For snarks, it guarantees that abnormal edges can be confined to a set of size at most \(\mu_3(G)\), which is always at most one fifth of the edge set [1911.06759].

## 5. Exact normal \(5\)-edge-colorings for structured snark families

Although the general \(5\)-color conjecture remains open, exact normal \(5\)-edge-colorings are known for several nontrivial snark families. One important example is a family of Loupekhine snarks built from Petersen blocks. For LP1-snarks, if the ring of blocks can be partitioned into three odd-length subsequences and every \(K_2\) component of the attachment graph \(G_E\) connects two vertices in the same subsequence, then the graph admits a normal \(5\)-edge-coloring. The construction colors the ring \(G_R\) so that every port sees colors \(1\) and \(2\), then colors the attachment edges of \(G_E\) by \(3\), \(4\), or \(5\) according to subsequence membership; edges of \(G_E\) become poor, while edges in the ring are checked to be poor or rich by local inspection. The same method extends to singly twisted LP2-snarks [1904.02661].

A distinct superposition framework replaces vertices and edges of a cycle in a snark by multipoles. For superpositions by Flower snarks, two sufficient conditions are established. If every superedge is dock-right, then the superposition admits a normal \(5\)-edge-coloring; more generally, if every superedge is both doubly-right and doubly-left, the same conclusion holds. Since Flower snarks are hypohamiltonian and the needed left/right colorings can be certified by suitable \(2\)-factors, this yields normal \(5\)-edge-colorings for all superpositions by Flower snarks considered there [2306.13340].

Superpositions by the Petersen graph exhibit a related but sharper parity phenomenon. For even cycles, a normal \(5\)-edge-coloring of the original snark extends through the superposition without changing colors outside the cycle, and the resulting coloring has at least \(18\) poor edges [2305.05981]. For odd cycles, such preservation can fail: the paper exhibits a \(5\)-cycle superposition of the Petersen graph where no normal \(5\)-edge-coloring extends the outside coloring unchanged [2305.05981]. A continuation resolves all remaining connection types for Petersen superpositions: if at least one local connection has \(p_i(1)\neq 1\) or \(d_i\neq 1\), then the superposed snark still admits a normal \(5\)-edge-coloring with at least \(18\) poor edges [2312.08739].

The broader landscape also includes full normal \(5\)-edge-colorings for several other named snark families and computational verification for all cubic graphs up to \(36\) vertices, as summarized in later work on partial normality [1911.06759]. This suggests that the main obstruction is not the existence of exact \(5\)-colorings in isolated constructions, but extending such constructions to arbitrary bridgeless cubic structure.

## 6. List normal edge-coloring and current open directions

The list variant replaces a fixed palette by edgewise lists. The list normal chromatic index \(ch'_N(G)\) is the smallest \(k\) such that every list assignment of size at least \(k\) admits a normal list edge-coloring. This parameter is substantially more restrictive than \(\chi'_N(G)\) [2410.01049].

For subcubic graphs,
\[
ch'_N(G)\le 10,
\]
because every strong edge-coloring is normal and the list strong chromatic index of subcubic graphs is at most \(10\) [2410.01049]. The lower bounds are already large: there is an infinite family of cubic graphs with \(ch'_N(G)\ge 9\), there exist bridgeless cubic graphs with \(ch'_N(G)\ge 8\), and there is an infinite family of cyclically \(4\)-edge-connected cubic graphs \(L_{2k}\) with \(ch'_N(L_{2k})\ge 7\) for \(k\ge 5\) [2410.01049].

These examples show that list normal edge-coloring can diverge sharply from ordinary normal edge-coloring. In particular, the bridgeless examples with \(ch'_N(G)\ge 8\) are remarked to be class I, hence \(\chi'_N(G)=3\) while \(ch'_N(G)\ge 8\) [2410.01049]. This is one of the clearest indications that local poor/rich constraints interact very differently with lists than with unrestricted palettes.

Several open problems organize the current state of the subject. The central question remains Jaeger’s conjecture: whether every bridgeless cubic graph satisfies \(\chi'_N(G)\le 5\) [1804.09449]. A related intermediate conjecture asks for \(\chi'_N(G)\le 6\) for all bridgeless cubic graphs; existing \(6\)-color theorems verify this only in substantial subclasses [1903.06043]. On the quantitative side, any improvement in the universal bound on \(\mu_3(G)\) would immediately improve the \(4/5\) guarantee for normal edges in proper \(5\)-edge-colorings [1911.06759]. On the structural side, extending the oddness-\(2\) theorem to oddness \(4\) appears difficult, and the recent authors explicitly note that they do not see a straightforward generalization beyond oddness \(2\) [2508.20565]. For the list theory, conjectured targets include \(ch'_N(G)\le 9\) for every cubic graph and \(ch'_N(G)\le 7\) for bridgeless subcubic graphs on at least \(14\) vertices [2410.01049].

Taken together, these results place normal edge-coloring in a distinctive position within cubic graph theory. It is simultaneously a local coloring constraint, a flow-theoretic phenomenon, a perfect-matching problem, and a reformulation of Petersen colorability. The mature part of the theory is the \(7\)-color and partial-normality regime; the frontier remains the transition from “almost normal” or “normal with six colors” to universal normal \(5\)-edge-colorability.

Source: https://www.emergentmind.com/topics/normal-edge-coloring