Palette Index and Palette Multigraph in Graphs

• The Reasoning Palette is the study of the palette index, which quantifies the diversity of incident color-sets in proper edge-colorings of multigraphs.
• Palette multigraphs encode repeated palette structures by translating vertex color profiles into a secondary graph that reflects connectivity and degree constraints.
• Recent constructions demonstrate that the palette index can grow quadratically with the maximum degree, setting a new extremal benchmark for edge-coloring problems.

The palette index is a chromatic parameter of a properly edge-colored graph or multigraph that measures the diversity of the local color-sets seen at vertices. For a loopless multigraph $G$ with a proper edge-coloring $c$, the palette of a vertex is the set of colors appearing on incident edges, and the palette index $\check s(G)$ is the minimum number of distinct such sets over all proper edge-colorings. The paper "A family of multigraphs with large palette index" develops a structural device called the palette multigraph and uses it to exhibit the first known family of multigraphs whose palette index is quadratic in the maximum degree $\Delta$, thereby establishing a quadratic lower benchmark for any universal upper bound in terms of $\Delta$ (Avesani et al., 2018).

1. Definition of palette index

Let $G$ be a loopless multigraph and let $c:E(G)\to C$ be a proper edge-coloring. The palette of a vertex $x$ is

$$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$

The palette index of $G$, denoted $c$0, is

$c$1

Thus the parameter asks for the smallest possible number of distinct local color-sets that can occur around vertices in any proper edge-coloring (Avesani et al., 2018).

This invariant is not a count of how many colors are used globally. Rather, it records how many different incident color profiles occur across the vertex set. Two vertices may have identical palettes even if the surrounding edge structure is different, provided they see the same set of colors. This makes the parameter sensitive to the interaction between edge-coloring and degree structure.

A useful point of comparison is that later work on the palette index of graphs emphasized the same basic definition and studied upper and lower bounds for ordinary graphs, especially bipartite and biregular families. That work also records basic facts such as $c$2 exactly for regular Class 1 graphs, and the impossibility of $c$3 for regular Class 2 graphs (Casselgren et al., 2018). Within that broader program, the 2018 multigraph construction isolates the phenomenon that palette index can be substantially larger in multigraphs than was previously witnessed.

2. Palette multigraph as a structural encoding

The central new object is the palette multigraph of an edge-colored multigraph $c$4, denoted $c$5. Its vertex set is the set of distinct palettes appearing in $c$6: $c$7

An edge $c$8 of $c$9 induces adjacency between the palette-vertices corresponding to $\check s(G)$0 and $\check s(G)$1. If $\check s(G)$2, then $\check s(G)$3 contains an ordinary edge between those two palette-vertices; if $\check s(G)$4, then it contributes a loop. Multiple edges and multiple loops are allowed when several edges of $\check s(G)$5 connect vertices with the same pair of palettes (Avesani et al., 2018).

This construction converts a coloring problem on $\check s(G)$6 into a multigraph on palette types. Two structural properties are highlighted. First, the number of edges of $\check s(G)$7, counted with multiplicity and loops, equals the number of edges of the underlying simple graph $\check s(G)$8. Second, Proposition 1 states that for any palette $\check s(G)$9, the degree of the corresponding vertex in $\Delta$0 is the sum of the degrees in $\Delta$1 of all vertices having that palette (Avesani et al., 2018).

The significance of this translation is that multiplicities of repeated palettes can be read off from degree information in $\Delta$2. The palette multigraph is therefore not merely a bookkeeping device: it links the number of vertices sharing a palette to graph-theoretic constraints such as acyclicity and average degree. In the main construction, that link is what produces a lower bound on the number of distinct palettes.

3. The family $\Delta$3 and its component pieces

The quadratic lower bound is obtained from a family of multigraphs $\Delta$4, indexed by the maximum degree $\Delta$5. The graph $\Delta$6 is defined as a disjoint union of components $\Delta$7 for

$\Delta$8

Each component $\Delta$9 is built from a windmill-like simple graph $\Delta$0 by replacing certain outer edges with $\Delta$1 parallel edges. In $\Delta$2, the central vertex $\Delta$3 has degree $\Delta$4, while every other vertex has degree $\Delta$5 (Avesani et al., 2018).

This degree pattern is the key combinatorial feature. Because the non-central vertices all have degree $\Delta$6, while the single central vertex has degree $\Delta$7, palettes arising at central and non-central vertices are constrained differently. Across different components $\Delta$8, vertices of different degrees cannot share the same palette. Consequently, when these components are assembled into the disjoint union $\Delta$9, the palette index across components accumulates almost additively, with the only possible overlap occurring among the $G$0-degree central vertices (Avesani et al., 2018).

A plausible implication is that the family is designed to separate local degree classes so sharply that any attempt to reuse palettes is forced into a very limited part of the graph. The lower bound is then driven by the impossibility of compressing palette types within each $G$1.

4. Forest structure after deleting the central palette

For an arbitrary proper edge-coloring $G$2 of $G$3, the argument studies the reduced palette multigraph

$G$4

obtained by deleting the palette-vertex corresponding to the central vertex $G$5 (Avesani et al., 2018).

Two facts are decisive.

First, $G$6 is a simple forest: it has no loops, no multiple edges, and no cycles. Second, the degree of a palette-vertex in $G$7 equals the number of vertices of $G$8 carrying that palette. This is Lemma 2, and it uses the fact that the non-central part of $G$9 consists of isolated edges in the underlying simple graph (Avesani et al., 2018).

Once the reduced palette multigraph is known to be a forest, the average degree is strictly less than $c:E(G)\to C$0. Because degree in $c:E(G)\to C$1 counts how many vertices share a palette, the average number of vertices per palette is also strictly less than $c:E(G)\to C$2. Since $c:E(G)\to C$3 has $c:E(G)\to C$4 vertices, every proper edge-coloring must therefore use more than $c:E(G)\to C$5 distinct palettes: $c:E(G)\to C$6

This is the core combinatorial mechanism. The lower bound does not come from counting colors or from a direct edge-coloring obstruction; it comes from forcing the quotient object $c:E(G)\to C$7 into a forest, where average-degree constraints limit how many original vertices can collapse onto the same palette type. This is why the paper identifies the special form of the palette multigraph as the key point of the argument (Avesani et al., 2018).

5. Quadratic lower bound in maximum degree

Summing the component-wise contribution across $c:E(G)\to C$8, the paper derives the main estimate

$c:E(G)\to C$9

Equivalently, the lower bound is

$x$0

so the palette index grows on the order of $x$1 (Avesani et al., 2018).

This was the first known family of multigraphs whose palette index is expressed by a quadratic polynomial in the maximum degree. The result has a direct extremal consequence: if there exists a universal polynomial upper bound $x$2 for the palette index of multigraphs, then that bound cannot be linear; it must be at least quadratic (Avesani et al., 2018).

The paper also notes a connected variant. By adding one extra vertex adjacent to every vertex of degree $x$3 in $x$4, one obtains a connected multigraph $x$5 of maximum degree $x$6 whose palette index still satisfies a quadratic lower bound in $x$7 (Avesani et al., 2018). This removes the possibility of attributing the phenomenon merely to disconnectedness.

A common misunderstanding is that large palette index should correlate directly with large chromatic index or with the number of colors used. The construction shows a different phenomenon: the obstruction is the number of distinct local color-sets that must appear, and that quantity can be forced to grow quadratically even when analyzed through proper edge-colorings alone.

6. Relation to other bounds, colorings, and open problems

The paper situates the result within a broader landscape of palette-index questions. It notes that previously known examples were much smaller in growth: complete graphs can have palette index bounded by a constant, and some trees exhibit growth of order $x$8, but no quadratic family had been known (Avesani et al., 2018). This places the construction at a new extremal scale.

The same paper also connects palette index to interval edge-colorings and interval cyclic edge-colorings. For multigraphs admitting interval edge-colorings, it proves the general upper bound

$x$9

This suggests that quadratic behavior is not an artifact of the lower-bound construction alone, but a natural magnitude in palette problems (Avesani et al., 2018).

Related work on ordinary graphs formulated the general conjecture that there exists a constant $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$0 such that

$$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$1

for every graph $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$2, and established this in several special settings, including complete multipartite graphs and many bipartite families (Casselgren et al., 2018). That later graph-theoretic evidence is consistent with the multigraph paper’s suggestion that a quadratic polynomial may suffice universally, while the family $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$3 shows that any such universal bound cannot be subquadratic (Avesani et al., 2018).

The resulting open problem is therefore sharply framed: whether every multigraph of maximum degree $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$4 satisfies $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$5 for some polynomial $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$6. The construction of $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$7 does not resolve that problem, but it establishes the lower threshold any successful answer must meet.

7. Conceptual importance of the palette multigraph method

The lasting conceptual contribution of the work is the introduction of the palette multigraph as an organizing object for palette-coloring structure. By turning repeated local color-sets into vertices of a secondary multigraph, the method isolates how palettes are distributed across adjacency relations and makes it possible to import structural arguments based on forests, degree sums, and average degree (Avesani et al., 2018).

This suggests a broader methodology for palette-index research. Instead of attacking $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$8 only through direct coloring constructions, one may analyze the geometry of the palette multigraph produced by an arbitrary proper edge-coloring. In the main family $$P_c(x)=\{\,c(e): e \text{ is incident with } x\,\}.$$9, deleting a single distinguished palette transforms the problem into the study of a simple forest; that reduction is what makes the lower bound transparent. A plausible implication is that future progress on universal palette-index bounds may depend on identifying other graph classes whose palette multigraphs admit similarly rigid structural descriptions.

The paper’s message is therefore twofold. At the extremal level, it proves that the palette index of multigraphs can genuinely be quadratic in $G$0. At the methodological level, it provides a structural language—the palette multigraph—for studying how local color-sets interact globally. Together, these two contributions define a natural agenda for subsequent work on palette index in graphs and multigraphs (Avesani et al., 2018).