arXiv is now an independent nonprofit! Learn more
License: CC BY-SA 4.0
arXiv:2608.19294v1 [math.CO] 19 Aug 2026

On Generalized Total Colourings of Planar Graphs

Philippe Cara Address: Department of Mathematics and Data Science
Vrije Universiteit Brussel
1050 Brussel
Belgium
Email address: philippe.cara@vub.be
and Samantha Dorfling Address: 1970,Belgium Email address: dorflingsamantha@gmail.com
Abstract.

In this paper we study generalised total colourings of graphs where the colour classes formed by vertices and edges, respectively, induce forests, while incident edges/vertices receive distinct colours. In [3] it was conjectured that for planar graphs, four colours suffice for this type of colouring. We confirm this conjecture for two infinite families of planar graphs.

Key words and phrases: 
generalized colouring, additive hereditary graph property, total colouring, maximal planar graph, planar graph, forest
2010 Mathematics Subject Classification
05C10, 05C15, 05C70

1. Introduction

In this paper we study simple, loopless, undirected graphs. For all undefined graph theoretical terms we refer the reader to [6]. For a graph GG, we will denote its set of vertices and its set of edges by V(G)V(G) and E(G)E(G), respectively.

The colourings we are interested in were introduced in the general context of hereditary properties, in [5]. Although we will only be concerned with one particular case of the so-called (𝒫,𝒬)({\cal P},{\cal Q})-total colourings, we briefly sketch the broader context for completeness, following [4]. Let {\cal I} denote the class of all graphs. A graph property is any isomorphism-closed subclass of {\cal I}. Let 𝒫{\cal P} be such a subclass. If a graph GG is a member of 𝒫{\cal P} we will say that GG has property 𝒫{\cal P}. A property 𝒫{\cal P} is called additive if for each graph GG, all of whose components have the property 𝒫\cal P, it follows that GG lies in 𝒫{\cal P} too. A property 𝒫{\cal P} is said to be hereditary if, whenever GG lies in 𝒫{\cal P}, and HH is a subgraph of GG, then HH also lies in 𝒫{\cal P}.

Well-known additive and hereditary graph properties are for example (see [4]):

  • 𝒪={G:E(G)=}{\cal O}=\{G\in{\cal I}:E(G)=\emptyset\},

  • 𝒮k={G:the maximum degree of G is at most k}{\cal S}_{k}=\{G\in{\cal I}:\mbox{the maximum degree of $G$ is at most $k$}\},

  • 𝒪k={G:each component of G has order at most k+1}{\cal O}_{k}=\{G\in{\cal I}:\mbox{each component of $G$ has order at most $k+1$}\},

  • 𝒲k={G:the length of the longest path in G is at most k}{\cal W}_{k}=\{G\in{\cal I}:\mbox{the length of the longest path in $G$ is at most $k$}\},

  • 𝒟k={G: every subgraph of G has minimum degree at most k}{\cal D}_{k}=\{G\in{\cal I}:\mbox{ every subgraph of $G$ has minimum degree at most $k$}\}.

Let 𝒫{\cal P} and 𝒬{\cal Q} denote additive hereditary graph properties. A (𝒫,𝒬)({\cal P},{\cal Q})-total colouring of a simple graph GG is a colouring of the vertices and edges of GG such that for each colour ii, the vertices coloured by ii induce a subgraph of GG with property 𝒫{\cal P}, the edges coloured by ii induce a subgraph of GG with property 𝒬{\cal Q}, and incident vertices and edges are coloured distinctly. The minimum number of colours needed such that a graph GG has a (𝒫,𝒬)({\cal P},{\cal Q})-total colouring is called the (𝒫,𝒬{\cal P},{\cal Q})-total chromatic number of GG and is denoted by χ𝒫,𝒬′′(G)\chi^{\prime\prime}_{{\cal P},{\cal Q}}(G).

Early studies of total colourings of graphs occur in [2], [1] and [9]. In [5], the authors then generalised the idea of total colourings of graphs and describe such colourings in the context of additive hereditary graph properties. In particular, they found upper and lower bounds for the generalized total colourings of graphs with various additive hereditary properties.

In [3], they study generalised total colourings for families of planar graphs and give upper bounds for these chromatic numbers for proper colourings of the vertices while the monochromatic edge sets are allowed to be forests. Using the terminology and symbols introduced above, the total chromatic number χ𝒪,𝒟1′′(G)\chi^{\prime\prime}_{{\cal O},{\cal D}_{1}}(G) is studied for GG planar. In particular they show that if an even planar triangulation has a Hamilton cycle HH for which there is no cycle among the edges inside HH, then such a graph needs at most four colours for a (𝒪,𝒟1)({\cal O},{\cal D}_{1})-total colouring as described above.

Furthermore, in [3], they conjecture that for all planar graphs GG one must have χ𝒟1,𝒟1′′(G)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(G)\leq 4. In this paper we define two classes of maximal planar graphs and show that for all graphs GG in these classes, χ𝒟1,𝒟1′′(G)=4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(G)=4. Therefore, we can confirm that for these two infinite classes of planar graphs the above-mentioned conjecture in [3] is true. We remark that, like in most results from [3], our graphs are hamiltonian.

2. Notation and basic results

A 𝒟1{\cal D}_{1} vertex colouring of a graph GG is a colouring of the vertices of GG such that the subgraphs of GG induced by these colour classes are forests (i.e. lie in 𝒟1{\cal D}_{1}).

A planar graph GG is called maximal planar if the addition of any edge to GG results in a nonplanar graph.

The following lemma is a well-known characterisation of maximal planar graphs, found in [6] for example:

Lemma 2.1.

A planar graph GG with order n3n\geq 3 and size mm is maximal planar if and only if m=3n6m=3n-6.

A maximal planar graph that contains a hamiltonian cycle will be called hamiltonian maximal planar.

We will construct a first class of maximal planar hamiltonian graphs.

Suppose that n4n\geq 4 is any integer. Let G=(V(G),E(G))G=(V(G),E(G)) be the planar graph we obtain as follows:
Let CnC_{n} denote a cycle on nn vertices; labelled v1,,vnv_{1},\ldots,v_{n}.
Add the following edges vivjv_{i}v_{j} to CnC_{n} (on the inside of CnC_{n}):
where i+j=n+2i+j=n+2 for all 2i,jn2\leq i,j\leq n and
where i+j=n+1i+j=n+1 for all 2i,jn2\leq i,j\leq n.
Finally, if n>4n>4, then also add the edges v1viv_{1}v_{i} for all 3i<n13\leq i<n-1 on the outside of the cycle. The class of all such graphs GG will be denoted by 𝖬𝖧\mathsf{MH}. Our construction is illustrated in Figure 1.

Refer to caption
Figure 1. The member of 𝖬𝖧\mathsf{MH} based on C12C_{12}.
Lemma 2.2.

Every graph in the class 𝖬𝖧\mathsf{MH} is maximal planar and hamiltonian.

Proof.

Let GG be any graph in 𝖬𝖧\mathsf{MH} with order n4n\geq 4. Then clearly GG is planar as well as Hamiltonian. From the construction above we obtain the size of GG in the case where nn is even and where nn is odd. First suppose that nn is odd, then to find the size of GG we add the nn cycle edges to the n32\frac{n-3}{2} inside edges vivjv_{i}v_{j} (where i+j=n+2i+j=n+2 and 2in2\leq i\leq n) and to this we add the n32\frac{n-3}{2} inside edges vivjv_{i}v_{j} (where i+j=n+1i+j=n+1 and 2in2\leq i\leq n) and then we add the n3n-3 outside edges v1viv_{1}v_{i} (where 3i<n13\leq i<n-1). Therefore the size of GG is n+n32+n32+n3=3n6n+\frac{n-3}{2}+\frac{n-3}{2}+n-3=3n-6 and thus, by Lemma 2.1, it follows that GG is maximal planar.

Next suppose that nn is even, then adding the edges in the same order as above gives size n+n22+n42+n3=3n6n+\frac{n-2}{2}+\frac{n-4}{2}+n-3=3n-6 and so again, by Lemma 2.1, GG is maximal planar. ∎

3. Generalized total colouring in the class 𝖬𝖧\mathsf{MH}

The main result in this section will show that for G𝖬𝖧G\in\mathsf{MH}, we have χ𝒟1,𝒟1′′(G)=4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(G)=4. In order to prove this, we will need a few lemmas.

First, it will be useful to define the following sets. For all k{0,1,2}k\in\{0,1,2\} let Sk={xxmod3=k}S_{k}=\{x\in\mathbb{N}\mid x\bmod 3=k\}.

Lemma 3.1.

Let GG be any graph in 𝖬𝖧\mathsf{MH} with vertex set V(G)={v1,v2,,vn}V(G)=\{v_{1},v_{2},\ldots,v_{n}\}. Then the vertex colouring c:V(G){1,2}c:V(G)\rightarrow\{1,2\} of GG, defined as c(v1)=1c(v_{1})=1 and for all iS2i\in S_{2} such that 2in22\leq i\leq\lceil\frac{n}{2}\rceil, c(vi)=c(vni)=1c(v_{i})=c(v_{n-i})=1 and otherwise c(vi)=2c(v_{i})=2, is a 𝒟1{\cal D}_{1} vertex colouring of GG.

Proof.

Let GG be any graph in 𝖬𝖧\mathsf{MH} with vertex set V(G)={v1,v2,,vn}V(G)=\{v_{1},v_{2},\ldots,v_{n}\} and let cc be the colouring in the statement of the lemma. For i=1,2i=1,2 we will use i\langle i\rangle to denote the set of all vertices vv in V(G)V(G) such that c(v)=ic(v)=i.

By the construction of graphs in 𝖬𝖧\mathsf{MH} and by definition of cc, we know that v1v_{1} is a universal vertex in GG and we know that c(v1)=1c(v_{1})=1. Furthermore, for all vertices viv_{i} and vjv_{j} in V(G)V(G) with c(vi)=c(vj)=1c(v_{i})=c(v_{j})=1 (i.e vi,vj1v_{i},v_{j}\in\langle 1\rangle), if i,j1i,j\neq 1, then vivjE(G)v_{i}v_{j}\not\in E(G): since otherwise, if vivjv_{i}v_{j} is an edge in GG, then, without loss of generality, 2in22\leq i\leq\lceil\frac{n}{2}\rceil and iS2i\in S_{2} while j=nkj=n-k for some kS2k\in S_{2} such that 2kn22\leq k\leq\lceil\frac{n}{2}\rceil. However, then i+j=(3l+2)+n(3m+2)=3(lm)+ni+j=(3l+2)+n-(3m+2)=3(l-m)+n for some integers ll and mm, and by the construction of graphs in 𝖬𝖧\mathsf{MH}, the edge vivjv_{i}v_{j} cannot exist. Therefore the subgraph of GG induced by the colour class 1\langle 1\rangle is acyclic.

Furthermore, the subgraph of GG induced by 2\langle 2\rangle is a path
P1:vn,vn1,v3,v4,vn3,vn4,v6,v7,vn6,vn7,,vn2P_{1}:v_{n},v_{n-1},v_{3},v_{4},v_{n-3},v_{n-4},v_{6},v_{7},v_{n-6},v_{n-7},\ldots,v_{\lceil\frac{n}{2}\rceil} in the case where nmod6{0,1,5}n\bmod 6\in\{0,1,5\} and a path P2:vn,vn1,v3,v4,vn3,vn4,v6,v7,vn6,vn7,,vn2+1P_{2}:v_{n},v_{n-1},v_{3},v_{4},v_{n-3},v_{n-4},v_{6},v_{7},v_{n-6},v_{n-7},\ldots,v_{\lceil\frac{n}{2}\rceil+1} in the case where nmod6{2,3,4}n\bmod 6\in\{2,3,4\}. Therefore the subgraph induced by the colour class 2\langle 2\rangle is also acyclic and thus the result holds. ∎

Refer to caption
Refer to caption
Figure 2. The vertex colouring and induced acyclic subgraphs in our example with 1212 vertices.

For every graph GG in 𝖬𝖧\mathsf{MH} with order n4n\geq 4, let GnG_{n} denote the graph obtained from GG as follows:

  1. (1)

    Colour the vertices of GG with the colouring cc defined in Lemma 3.1.

  2. (2)

    Remove all edges from GG that join vertices with the same colour.

The graph G12G_{12} obtained from our example of order 1212 is shown in Figure 3.

Refer to caption
Figure 3. The subgraph G12G_{12}.

Note that the graph GnG_{n} is planar and that every edge of GnG_{n} lies on a cycle of length four. Furthermore, for all integers n4n\geq 4, the graph GnG_{n} contains a vertex of degree 2, and in particular, deg(vn)Gn=2{}_{G_{n}}(v_{n})=2.

In the following lemma, let GnG_{n} denote the graph constructed above, with vertex set V(G)={v1,v2,,vn}V(G)=\{v_{1},v_{2},\ldots,v_{n}\}, numbered as in the construction above.

Lemma 3.2.

For all positive integers n5n\geq 5, if HH and GG are graphs in 𝖬𝖧\mathsf{MH} with vertex sets V(H)={u1,u2,,un1}V(H)=\{u_{1},u_{2},\ldots,u_{n-1}\} and V(G)={v1,v2,,vn}V(G)=\{v_{1},v_{2},\ldots,v_{n}\} respectively, then the graph Hn1H_{n-1} is isomorphic to Gnvn+12G_{n}-v_{\lceil\frac{n+1}{2}\rceil} if nn is even, and isomorphic to Gnvn2G_{n}-v_{\lceil\frac{n}{2}\rceil} if nn is odd. In both cases the removed vertex has degree 22 in GnG_{n}.

Proof.

Take any integer n5n\geq 5 and let the graph Hn1H_{n-1} be one that is constructed from a graph HH in 𝖬𝖧\mathsf{MH}, with order n1n-1 and V(H)={u1,u2,,un1}V(H)=\{u_{1},u_{2},\ldots,u_{n-1}\}. By Lemma 2.2, HH is a triangulation and we may choose an embedding in the plane such that the vertices u1,un12u_{1},u_{\lceil\frac{n-1}{2}\rceil} and un12+1u_{\lceil\frac{n-1}{2}\rceil+1} lie on the outer face of HH. Note that any graph GG in 𝖬𝖧\mathsf{MH} (up to isomorphism) with order nn, can be obtained from HH by joining a single vertex vv with the vertices u1,un12u_{1},u_{\lceil\frac{n-1}{2}\rceil} and un12+1u_{\lceil\frac{n-1}{2}\rceil+1} in HH. One can then easily see the correspondence between the vertices V(G)={v1,v2,,vn}V(G)=\{v_{1},v_{2},\ldots,v_{n}\} used in the construction of GG and the vertices {u1,u2,,un12,v,un12+1,,un1}=V(H){v}\{u_{1},u_{2},\ldots,u_{\lceil\frac{n-1}{2}\rceil},v,u_{\lceil\frac{n-1}{2}\rceil+1},\ldots,u_{n-1}\}=V(H)\cup\{v\}.

Furthermore, if we colour the vertices of GG with the colouring cc defined in Lemma 3.1, then since this colouring is acyclic and vn12vn12+2v_{\lceil\frac{n-1}{2}\rceil}v_{\lceil\frac{n-1}{2}\rceil+2} is an edge in GG (corresponding to the edge un12un12+1u_{\lceil\frac{n-1}{2}\rceil}u_{\lceil\frac{n-1}{2}\rceil+1} in HH), that forms a cycle of length 33 together with the vertex v1v_{1} which has colour 11, at least one of the vertices vn12v_{\lceil\frac{n-1}{2}\rceil} and vn12+2v_{\lceil\frac{n-1}{2}\rceil+2} must have colour 22. As the colouring is acyclic, the case c(vn12)=c(vn12+2)=2c(v_{\lceil\frac{n-1}{2}\rceil})=c(v_{\lceil\frac{n-1}{2}\rceil+2})=2 will yield c(vn12+1)=1c(v_{\lceil\frac{n-1}{2}\rceil+1})=1, implying that the degree of vn12+1v_{\lceil\frac{n-1}{2}\rceil+1} will be 22 in GnG_{n}. In the case that vn12v_{\lceil\frac{n-1}{2}\rceil} and vn12+2v_{\lceil\frac{n-1}{2}\rceil+2} have different colours, the cycle consisting of v1v_{1}, vn12+1v_{\lceil\frac{n-1}{2}\rceil+1} and the vertex of colour 11 in {vn12,vn12+2}\{v_{\lceil\frac{n-1}{2}\rceil},v_{\lceil\frac{n-1}{2}\rceil+2}\} will yield colour 22 for vn12+1v_{\lceil\frac{n-1}{2}\rceil+1}, showing again that the degree of vn12+1v_{\lceil\frac{n-1}{2}\rceil+1} is 22 in GnG_{n}.

Now, if vn12+1v_{\lceil\frac{n-1}{2}\rceil+1} is removed from the graph GnG_{n}, then the only edges lost, come from the set {vn12vn12+1,vn12+1vn12+2,v1vn12+1}\{v_{\lceil\frac{n-1}{2}\rceil}v_{\lceil\frac{n-1}{2}\rceil+1},v_{\lceil\frac{n-1}{2}\rceil+1}v_{\lceil\frac{n-1}{2}\rceil+2},v_{1}v_{\lceil\frac{n-1}{2}\rceil+1}\} and the remaining graph is isomorphic to Hn1H_{n-1}, as the colouring of HH is consistent with the colouring of GG constructed from HH by adding the vertex vv.

To end the proof we remark that if nn is even we have n12+1=n2+1\lceil\frac{n-1}{2}\rceil+1=\lceil\frac{n}{2}\rceil+1 and that if nn is odd we have n12+1=n2\lceil\frac{n-1}{2}\rceil+1=\lceil\frac{n}{2}\rceil. So the vertex vn12+1v_{\lceil\frac{n-1}{2}\rceil+1} that we remove from GnG_{n} is indeed equal to either vn2+1v_{\lceil\frac{n}{2}\rceil+1} or vn2v_{\lceil\frac{n}{2}\rceil}, depending on the parity of nn. ∎

Lemma 3.3.

Let n4n\geq 4 be any positive integer and GG be any graph in 𝖬𝖧\mathsf{MH} with vertex set V(G)={v1,v2,,vn}V(G)=\{v_{1},v_{2},\ldots,v_{n}\}. Then the edge set of the graph GnG_{n} can be decomposed into two sets F1F_{1} and F2F_{2} such that the graphs induced by each of these sets lie in 𝒟1{\cal D}_{1}.

Proof.

Let GnG_{n} be the graph constructed from GG with vertex set V(Gn)=V(G)={v1,v2,,vn}V(G_{n})=V(G)=\{v_{1},v_{2},\ldots,v_{n}\}. From Lemma 3.2, there exists a vertex vV(Gn)v\in V(G_{n}) such that deg(v)=2(v)=2, in particular, if nn is odd, then one can take v=vn2v=v_{\lceil\frac{n}{2}\rceil} and if nn is even, then v=vn+12v=v_{\lceil\frac{n+1}{2}\rceil} has degree 22. To decompose the edges of GnG_{n} as required, we consider two copies F1F_{1} and F2F_{2} of the empty graph nK1nK_{1}, with the vertices of each labelled as in GnG_{n}. Consider the graph GnG_{n}. We now apply the following algorithm:

Place one of the two edges incident with vv (in GnG_{n}), between the corresponding vertices in F1F_{1} and the other edge between the corresponding vertices in F2F_{2}.
Remove the vertex vv from GnG_{n}. Note that, by Lemma 3.2, the resulting graph GnvG_{n}-v is isomorphic to a graph Gn1G_{n-1} where Gn1G_{n-1} contains a vertex uu such that deg(u)=2(u)=2.
Continue with the above algorithm until the resulting graph is isomorphic to a cycle G4C4G_{4}\cong C_{4} of length four.
Let xx denote any vertex in G4G_{4}. Place one of the two edges incident with xx (in G4G_{4}) between the corresponding vertices in F1F_{1} and the other edge between the corresponding vertices in F2F_{2}.
Remove xx from G4G_{4}.
Finally place one of the two remaining edges in the resulting G2G_{2} (P3)(\cong P_{3}) between the corresponding vertices in F1F_{1} and the other edge between the corresponding vertices in F2F_{2}.

By the construction of F1F_{1} and F2F_{2} above it is easy to see that the graphs induced by both F1F_{1} and F2F_{2} are acyclic, since once the edge incident with a particular vertex vv, in some graph GiG_{i} (2in2\leq i\leq n), such that deg(v)=2(v)=2, is added to FjF_{j}, (j{1,2}j\in\{1,2\}), the vertex vv is never again encountered in FjF_{j} as it is deleted from GiG_{i}.

Using the above, we are able to decompose the edge set of GnG_{n} into two sets such that the graph induced by edges in F1F_{1} and the graph induced by edges in F2F_{2} both lie in 𝒟1{\cal D}_{1}. ∎

Refer to caption
Refer to caption
Refer to caption
Figure 4. The decomposition of G12G_{12} into two forests.
Theorem 3.4.

For all integers n4n\geq 4, every graph GG in 𝖬𝖧\mathsf{MH} with order nn satisfies χ𝒟1,𝒟1′′(G)=4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(G)=4.

Proof.

Let GG be any graph in 𝖬𝖧\mathsf{MH} with order n4n\geq 4 and vertex set V(G)={v1,v2,,vn}V(G)=\{v_{1},v_{2},\ldots,v_{n}\}. Colour the vertices of GG with the colouring c:V(G){1,2}c:V(G)\rightarrow\{1,2\} of V(G)V(G) such that c(v1)=1c(v_{1})=1 and for all iS2i\in S_{2} such that 2in22\leq i\leq\lceil\frac{n}{2}\rceil, c(vi)=c(vni)=1c(v_{i})=c(v_{n-i})=1 and otherwise c(vi)=2c(v_{i})=2. By Lemma 3.1, the subgraphs induced by the two colour classes 1\langle 1\rangle and 2\langle 2\rangle of cc are acyclic. Consider the subgraph GnG_{n} of GG. Applying the method used in the proof of Lemma 3.3, colour the edges of GnG_{n} induced by F1F_{1} with colour 3 and the edges induced by F2F_{2} with colour 4. Finally, for all edges vivjGv_{i}v_{j}\in G such that vivjGnv_{i}v_{j}\not\in G_{n}, (1i,jn1\leq i,j\leq n) if c(vi)=c(vj)=1c(v_{i})=c(v_{j})=1, then colour the edge vivjv_{i}v_{j} with colour 2, and for all edges vivjGv_{i}v_{j}\in G such that vivjGnv_{i}v_{j}\not\in G_{n}, (1i,jn1\leq i,j\leq n) if c(vi)=c(vj)=2c(v_{i})=c(v_{j})=2, colour the edge vivjv_{i}v_{j} with colour 1.

The total colouring above provides a 𝒟1{\cal D}_{1} colouring of the vertices of GG as well as a 𝒟1{\cal D}_{1} colouring of the edges of GG such that incident vertices and edges receive distinct colours and therefore
χ𝒟1,𝒟1′′(G)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(G)\leq 4.

Furthermore, the complete graph K4K_{4} is the smallest graph in 𝖬𝖧\mathsf{MH} and clearly K4K_{4} is a subgraph of all graphs in 𝖬𝖧\mathsf{MH}. However, if we try to find a total colouring of K4K_{4}, then we see that, firstly, we need at least two colours for the vertices and since K4K_{4} is a triangulation, at most two vertices may have the same colour. Thus if we use exactly two colours for the vertices, then there are two vertices of each colour. However, then since K4K_{4} is complete, we can always find a 4-cycle C:u1,u2,u3,u4,u1C:u_{1},u_{2},u_{3},u_{4},u_{1} in the graph with vertices of alternating colours. However, then none of the edges between the vertices on CC may receive colour 1 or 2. Furthermore, since CC is a cycle, we will need an additional two colours to colour its edges. However, if we use three colours for the vertices of K4K_{4}, then two vertices must have the same colour, say 1. Call the other two vertices uu and vv and call their colours 22 and 33 respectively. Then the only edge that may be coloured with 11 is uvuv. Now the remaining five edges are the following: the edge between the two vertices of colour 11, two vertices between uu and the vertices with colour 11 and two between vv and the vertices with colour 11. They form two cycles of length 33 sharing the edge between the vertices of colour 11. Trying to use only 33 colours, the edges on uu must be coloured with colour 33 and the ones on vv with colour 22. For the edge between the two vertices of colour 11 either of the colours 22 or 33 will yield a monochromatic cycle of length 33 so we will need a fourth colour. Therefore χ𝒟1,𝒟1′′(K4)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(K_{4})\geq 4. Since K4K_{4} is a subgraph of every graph in 𝖬𝖧\mathsf{MH}, it follows that χ𝒟1,𝒟1′′(G)=4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(G)=4. ∎

The following corollary to Theorem 3.4 follows from [5], since for a positive integer kk, a subgraph HH of a graph GG and additive hereditary properties 𝒫{\cal P} and 𝒬{\cal Q}, if χ𝒫,𝒬′′(G)k\chi^{\prime\prime}_{{\cal P},{\cal Q}}(G)\leq k, then χ𝒫,𝒬′′(H)k\chi^{\prime\prime}_{{\cal P},{\cal Q}}(H)\leq k.

Corollary 3.5.

Every (planar) subgraph of a graph GG in 𝖬𝖧\mathsf{MH} satisfies
χ𝒟1,𝒟1′′(G)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(G)\leq 4.

4. Generalized total colourings for triangulated square grids

We now construct another family of hamiltonian maximal planar graphs.

Let k2k\geq 2 and l2l\geq 2 denote positive integers and PkP_{k} and PlP_{l} denote paths of order kk and ll, respectively. The cartesian product of PkP_{k} and PlP_{l} is called a k×lk\times l grid or a k×lk\times l lattice graph and has order klkl.

Let GG be a n×nn\times n grid for some positive integer n3n\geq 3 and let V(G)V(G) denote the vertex set of GG. To label the vertices of a square grid GG, we will use notation similar to that used for entries of a matrix. Thus vijv_{ij} will denote the vertex in row ii and column jj of our grid. We will define a triangulated grid of GG to be the graph GtGt obtained from GG by adding a diagonal edge from the top right corner to the bottom left corner of every induced subgraph HH of GG such that HH is isomorphic to a cycle of length four, and we will add an edge between the vertex v11v_{11} and every vertex on the outer face of V(G)V(G).

More formally, if V(G)={v11,v12,,vnn}V(G)=\{v_{11},v_{12},\ldots,v_{nn}\}, then GtGt is the graph obtained by adding all edges vijv(i+1)(j1)v_{ij}v_{(i+1)(j-1)} (for all 1i,jn1\leq i,j\leq n) and all edges v11v1jv_{11}v_{1j}, v11vi1v_{11}v_{i1}, v11vnjv_{11}v_{nj} and v11vinv_{11}v_{in} for all 1in11\leq i\leq n-1 and 2jn2\leq j\leq n to GG. It is easy to see that the graph GtGt is planar. Furthermore, the graph GtGt is maximal planar, since it has order n2n^{2} and size 2(n1)n+(n1)2+(4n43)=3n262(n-1)n+(n-1)^{2}+(4n-4-3)=3n^{2}-6 (adding edges on the grid GG to the diagonal edges and adding this to the order of the neighbourhood of v11v_{11} without the edges v12v_{12} and v21v_{21}).

Refer to caption
Figure 5. The triangulated 4×44\times 4 grid.
Theorem 4.1.

Let GG be a n×nn\times n grid for any integer n3n\geq 3. Then the triangulated grid GtGt satisfies, χ𝒟1,𝒟1′′(Gt)=4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(Gt)=4.

Proof.

First we will show that χ𝒟1,𝒟1′′(Gt)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(Gt)\leq 4. Similar to the proof of Theorem 3.4, in order to show this upper bound, we will first define a vertex colouring of a graph GtGt that will use two colours and we will show that the graphs induced by the two colour classes are both acyclic. Then we will show that the graph HH resulting from GtGt by removing all edges whose end vertices lie in the same colour class can be edge-partitioned into two spanning forests F1F_{1} and F2F_{2}. From this we will be able to provide the necessary total colouring.

Take any graph isomorphic to a triangulated grid GtGt.

Consider the vertex colouring c:V(Gt){1,2}c:V(Gt)\rightarrow\{1,2\} defined by

c(vij)={1 if 1ij and i+j is even, or if 1ji3,i4 and i+j is odd, and 2 otherwise. c(v_{ij})=\left\{\begin{array}[]{ll}1&\mbox{ if }1\leq i\leq j\mbox{ and }i+j\mbox{ is even, or}\\ &\mbox{ if }1\leq j\leq i-3,i\geq 4\mbox{ and }i+j\mbox{ is odd, and }\\ 2&\mbox{ otherwise. }\end{array}\right.

Let 1\langle 1\rangle and 2\langle 2\rangle denote the colour classes of cc, then the graph induced by the vertices in V(Gt)V(Gt) that lie in 1\langle 1\rangle, is acyclic and the same holds for the graph induced by the vertices that lie in 2\langle 2\rangle: this is easily seen, since by definitions of the edge set E(Gt)E(Gt) of GtGt and the colouring cc, the graph (say G2G_{2}) induced on the vertices in 2\langle 2\rangle is a generalized caterpillar with, as spine, the path v12,v21,v31,,v(n1)nv_{12},v_{21},v_{31},\ldots,v_{(n-1)n} of length 2n22n-2 and legs of varying lengths kk such that 0kn20\leq k\leq\lceil\frac{n}{2}\rceil. Furthermore, the graph (say G1G_{1}) induced on the vertices of 1\langle 1\rangle is a generalized star with center v11v_{11} and with 2n42n-4 arms of varying lengths kk such that 1kn21\leq k\leq\lceil\frac{n}{2}\rceil.

Now let HH denote the spanning subgraph of GtGt that results when we remove the edge sets of G1G_{1} and G2G_{2} from GtGt.

Let F1F_{1} denote the subgraph of HH induced by the edges {vijvklE(H)|i=k1,j=l,c(vij)=1 and c(vkl)=2}{vijvklE(H)|i=k,j=l1,c(vij)=1 and c(vkl)=2}{vijvklE(H)|i=k1,j=l+1,c(vij)=1 and c(vkl)=2}{v11vklE(H)|k=1 and c(vkl)=2, or l=n and c(vkl)=2}\{v_{ij}v_{kl}\in E(H)\ |\ i=k-1,j=l,c(v_{ij})=1\mbox{ and }c(v_{kl})=2\}\cup\{v_{ij}v_{kl}\in E(H)\ |\ i=k,j=l-1,c(v_{ij})=1\mbox{ and }c(v_{kl})=2\}\cup\{v_{ij}v_{kl}\in E(H)\ |\ i=k-1,j=l+1,c(v_{ij})=1\mbox{ and }c(v_{kl})=2\}\cup\{v_{11}v_{kl}\in E(H)\ |\ k=1\mbox{ and }c(v_{kl})=2,\mbox{ or }l=n\mbox{ and }c(v_{kl})=2\}. It is easily seen that F1F_{1} is a forest (a generalized star with v11v_{11} its central vertex and with nn arms, such that every arm is a tree).

Let F2F_{2} denote the subgraph of HH induced by the edges {vijvklE(H)|i=k1,j=l,c(vij)=2 and c(vkl)=1}{vijvklE(H)|i=k,j=l1,c(vij)=2 and c(vkl)=1}{vijvklE(H)|i=k1,j=l+1,c(vij)=2 and c(vkl)=1}{v11vklE(H)|l=1 and c(vkl)=2, or k=n and c(vkl)=2}\{v_{ij}v_{kl}\in E(H)\ |\ i=k-1,j=l,c(v_{ij})=2\mbox{ and }c(v_{kl})=1\}\cup\{v_{ij}v_{kl}\in E(H)\ |\ i=k,j=l-1,c(v_{ij})=2\mbox{ and }c(v_{kl})=1\}\cup\{v_{ij}v_{kl}\in E(H)\ |\ i=k-1,j=l+1,c(v_{ij})=2\mbox{ and }c(v_{kl})=1\}\cup\{v_{11}v_{kl}\in E(H)\ |\ l=1\mbox{ and }c(v_{kl})=2,\mbox{ or }k=n\mbox{ and }c(v_{kl})=2\}. It is easily seen that F2F_{2} is a forest (the union of a generalized star with n1n-1 arms (such that the arms are paths) together with a union of paths).

Therefore, to show that χ𝒟1,𝒟1′′(Gt)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(Gt)\leq 4, we start with a graph GtGt and we colour its vertices with the colouring cc defined above to obtain two colour classes 1\langle 1\rangle and 2\langle 2\rangle. To the edges of GtGt, first we colour the edges of the graph G1G_{1} (defined above) with the colour 2 and we colour the edges of the graph G2G_{2} with the colour 1. Next we colour the edges of the forest F1F_{1} with a new colour 3 and finally the edges of the forest F2F_{2} with a new colour 4. From our observations above, we have that the subgraphs induced by vertices with the same colour lie in 𝒟1{\cal D}_{1}, all incident vertices and edges have distinct colours and the graphs induced by edges with the same colour lie in 𝒟1{\cal D}_{1}. Thus we may conclude that χ𝒟1,𝒟1′′(Gt)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(Gt)\leq 4.

Now we will show that χ𝒟1,𝒟1′′(Gt)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(Gt)\geq 4. Note that the induced subgraph v11,v1(n1),v1n,v2(n1),v2n\langle v_{11},v_{1(n-1)},v_{1n},v_{2(n-1)},v_{2n}\rangle of any triangulated grid GtG_{t} is isomorphic to a wheel, which we will denote by WW, which has order 5 and central vertex v1nv_{1n}. We will show that χ𝒟1,𝒟1′′(W)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(W)\geq 4, and therefore assume (to the contrary) that χ𝒟1,𝒟1′′(W)3\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(W)\leq 3. Since WW contains cycles, we need at least two colours with which to colour the vertex set of WW. For the sake of simplicity, let’s label the vertices of WW as {u1,u2,u3,u4,x}\{u_{1},u_{2},u_{3},u_{4},x\}, where xx denotes the central vertex of WW. We begin by colouring the vertices in WW that lie on the outer 4-cycle (which we will denote by CC), i.e., all vertices excepting the vertex xx. Consider the total colouring c:{u1,u2,u3,u4,x}E(W){1,2,3}c:\{u_{1},u_{2},u_{3},u_{4},x\}\cup E(W)\rightarrow\{1,2,3\} of V(W)E(W)V(W)\cup E(W).
Case 1: Suppose that c(u1)=c(u3)c(u2)=c(u4)c(u_{1})=c(u_{3})\neq c(u_{2})=c(u_{4}). Then, since incident vertices and edges must receive distinct colours and CC is a cycle, we will need two new colours to colour E(C)E(C). Therefore 4χ𝒟1,𝒟1′′(C)χ𝒟1,𝒟1′′(W)34\leq\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(C)\leq\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(W)\leq 3 — a contradiction.
Case 2: Suppose now that c(u1)=c(u4)c(u2)=c(u3)c(u_{1})=c(u_{4})\neq c(u_{2})=c(u_{3}) and, without loss of generality, c(u1)=1c(u_{1})=1 and c(u2)=2c(u_{2})=2. Then, in WW, the vertex xx must satisfy c(x)=3c(x)=3. In order to satisfy the restrictions imposed by the total colouring, we know that c(u1x)=c(u4x)=2c(u_{1}x)=c(u_{4}x)=2 and c(u2x)=c(u3x)=1c(u_{2}x)=c(u_{3}x)=1. However, then c(u1u4)=3c(u_{1}u_{4})=3 and thus the only available colour assignment for u2u3u_{2}u_{3} is c(u2u3)=3c(u_{2}u_{3})=3 — now all the edges of CC are coloured the same, a contradiction.
Case 3: Finally, suppose that c(u1)=c(u2)=c(u3)c(u4)c(u_{1})=c(u_{2})=c(u_{3})\neq c(u_{4}) and (without loss of generality) that c(u1)=1c(u_{1})=1 and c(u4)=2c(u_{4})=2. Clearly, c(x)1c(x)\neq 1.
If c(x)=2c(x)=2, then c(u1x)=c(u3x)=c(u3u4)=c(u1u4)=3c(u_{1}x)=c(u_{3}x)=c(u_{3}u_{4})=c(u_{1}u_{4})=3 and thus the colour class 3\langle 3\rangle induces a cycle u1,x,u3,u4u_{1},x,u_{3},u_{4} — a contradiction. Thus c(x)=3c(x)=3. However, this forces c(u1u4)=c(u3u4)=3c(u_{1}u_{4})=c(u_{3}u_{4})=3 and c(u2x)=c(u3x)=2c(u_{2}x)=c(u_{3}x)=2 and the latter colouring forces c(u2u3)=3c(u_{2}u_{3})=3. This makes it impossible for us to colour the edge u1u2u_{1}u_{2} with one of the three colours since c(u1u2)1c(u_{1}u_{2})\neq 1 as c(u1)=c(u2)=1c(u_{1})=c(u_{2})=1; also c(u1u2)=3c(u_{1}u_{2})=3 causes 3\langle 3\rangle to induce a cycle isomorpic to CC and finally c(u1u2)=2c(u_{1}u_{2})=2 results in a triangle u1,u2,xu_{1},u_{2},x induced by 2\langle 2\rangle.
Therefore χ𝒟1,𝒟1′′(W)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(W)\geq 4. Since every triangulated square grid GtGt contains the wheel WW, we may conclude that χ𝒟1,𝒟1′′(Gt)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(Gt)\geq 4 from which the result follows. ∎

Refer to caption
Refer to caption
Figure 6. The vertex colouring and induced acyclic subgraphs in the triangulated 4×44\times 4 grid.
Refer to caption
Refer to caption
Refer to caption
Figure 7. The decomposition of the triangulated 4×44\times 4 grid into two forests.
Corollary 4.2.

Let GG be a n×nn\times n grid for any integer n3n\geq 3 and let GtGt denote the triangulated grid of GG. Then every (planar) subgraph of GtGt satisfies χ𝒟1,𝒟1′′(Gt)4\chi^{\prime\prime}_{{\cal D}_{1},{\cal D}_{1}}(Gt)\leq 4.

References

  • [1] M. Behzad, The total chromatic number of a graph, a survey, Proc. Conference Combinatorial Mathematics (Oxford, England 1969), Academic Press New York (1970).
  • [2] M. Behzad, G. Chartrand, J.K. Cooper Jr, The numbers of complete graphs, J. London. Math. Society, 42(1967) 226–228.
  • [3] M. Borowiecki and I. Broere, Hamiltonicity and Generalised Total Colourings of planar graphs, Discussiones Mathematicae Graph Theory 36 (2016) 243–257.
  • [4] M. Borowiecki, I. Broere, M. Frick, P. Mihók and G. Semanišin, A survey of hereditary properties of graphs, Discussiones Mathematicae Graph Theory, 17 (1997) 5–50.
  • [5] M. Borowiecki, A. Kemnitz and P. Mihók, Generalized total Colourings of graphs, Discussiones Mathematicae Graph Theory, 31 (2011) 209–222.
  • [6] G. Chartrand, L. Lesniak and P. Zhang, Graphs & Digraphs, fifth edition, Chapman & Hall/CRC, 2010, ISBN 1-43982-627-7.
  • [7] G. Karafová, Generalized Fractional Total Colorings of Complete Graphs, Discussiones Mathematicae Graph Theory, 33 (4) (2013) 665–676.
  • [8] J.B. Kruskal, On the shortest spanning tree of a graph and the traveling salesman problem, Proc. Amer. Math. Soc., 7 (1956) 48–50.
  • [9] N. Vijayaditya, On total chromatic number of a graph, Journal of the London Mathematical Society (2), 3 (1971), 405–408.