Discrete Applied Mathematics 184 (2015) 189–195

Contents lists available at ScienceDirect

Discrete Applied Mathematics

journal homepage: www.elsevier.com/locate/dam

AVD-total-colouring of complete equipartite graphs✩

Atílio G. Luiz ∗ , C.N. Campos, C.P. de Mello
Institute of Computing, University of Campinas, Brazil

article info abstract

Article history: An AVD-total-colouring of a simple graph G is a mapping π : V (G) ∪ E (G) → C , C a set
Received 4 June 2013 of colours, such that: (i) for each pair of adjacent or incident elements x, y ∈ V (G) ∪ E (G),
Received in revised form 1 November 2014 π (x) ̸= π (y); (ii) for each pair of adjacent vertices x, y ∈ V (G), sets {π (x)} ∪ {π (xv): xv ∈
Accepted 11 November 2014
E (G), v ∈ V (G)} and {π (y)} ∪ {π (yv): yv ∈ E (G), v ∈ V (G)} are distinct. The AVD-
Available online 19 January 2015
total-chromatic number, χa′′ (G), is the smallest number of colours for which G admits an
AVD-total-colouring. In 2005, Zhang et al. conjectured that χa′′ (G) ≤ ∆(G) + 3 for any
Keywords:
Total-colouring
simple graph G. In this article this conjecture is verified for any complete equipartite
Adjacent-vertex-distinguishing-total- graph. Moreover, if G is a complete equipartite graph of even order, then it is shown that
colouring χa′′ (G) = ∆(G) + 2.
Complete equipartite graph © 2014 Elsevier B.V. All rights reserved.

1. Introduction

A graph G is defined by a set V (G) of elements called vertices, a set E (G) of elements called edges, and a relation of incidence,
which associates with each edge either one or two vertices called its ends. An edge is called a link or a loop according as the
number of its ends is two or one, respectively. Two or more links with the same pair of ends are called multiple edges. A
graph without loops and multiple edges is called a simple graph. The cardinality of V (G) is the order of G. We denote an edge
e ∈ E (G) by uv when u and v are its ends. An element of G is a vertex or an edge of G. As usual, we denote by d(v) the degree
of a vertex v ∈ V (G), and by ∆(G) the maximum degree of G.
Let C be a set of colours. A vertex-colouring of G is a mapping ϕ : V (G) → C such that for each pair of adjacent vertices
u, v ∈ V (G), we have ϕ(u) ̸= ϕ(v). Similarly, an edge-colouring of G is a mapping λ: E (G) → C such that for each pair of
adjacent edges uv, vw ∈ E (G), we have λ(uv) ̸= λ(vw). A total-colouring of G is a mapping π : V (G) ∪ E (G) → C such
that for each pair of adjacent or incident elements x, y ∈ V (G) ∪ E (G), we have π (x) ̸= π (y). If |C | = k, then π is called a
k-total-colouring of G, λ is called a k-edge-colouring of G, and ϕ is called a k-vertex-colouring of G.
The chromatic number of G, denoted χ (G), is the smallest number of colours for which G admits a vertex-colouring.
Similarly, the chromatic index of G, denoted χ ′ (G), is the smallest number of colours for which G admits an edge-colouring;
and the total-chromatic number of G, denoted χ ′′ (G), is the smallest number of colours for which G admits a total-colouring.
Let π be a total-colouring of G and let Cπ (u) := {π (u)} ∪ {π (uv) : uv ∈ E (G), v ∈ V (G)} be the set of colours that occur
in a vertex u ∈ V (G). If it is clear from the context that π is a total-colouring of G, then Cπ (u) is written simply as C (u). We
denote by C (u) the set of colours of C that do not occur in vertex u ∈ V (G). Two vertices, u and v , are distinguishable when
C (u) ̸= C (v). If this property is true for every pair of adjacent vertices, then π is an Adjacent-Vertex-Distinguishing-Total-
Colouring (AVD-total-colouring). The AVD-total-chromatic number of G, denoted χa′′ (G), is the smallest number of colours for
which G admits an AVD-total-colouring. If |C | = k, we say that π uses k colours, and π is called a k-AVD-total-colouring.

✩ This research was supported by CNPq and FAPESP.

∗ Corresponding author.
E-mail addresses: [email protected] (A.G. Luiz), [email protected] (C.N. Campos), [email protected] (C.P. de Mello).

http://dx.doi.org/10.1016/j.dam.2014.11.006
0166-218X/© 2014 Elsevier B.V. All rights reserved.
190 A.G. Luiz et al. / Discrete Applied Mathematics 184 (2015) 189–195

(a) K (4 × 2) endowed with canonical (b) Two disjoint induced

labelling. subgraphs of K (4 × 2)
isomorphic to K4 .

Fig. 1. K (4 × 2) and its induced subgraphs K 1 and K 2 that are isomorphic to K4 .

For an edge-colouring, set C (u) is defined by the colours of the edges incident with u. A vertex-distinguishing-proper-edge-
colouring is an edge-colouring of G that requires C (u) ̸= C (v) for each u, v ∈ V (G), u ̸= v . This colouring was first examined
by Burris and Schelp [3], and further investigated by others, including Bazgan et al. [2] and Balister et al. [1]. The motivation
for studying vertex-distinguishing-proper-edge-colourings came from irregular networks [4]. In an irregular network, each
edge is assigned a positive integer weight and each vertex is assigned the sum of the weights of its incident edges, in such
a way that any two vertices receive distinct values. Zhang et al. [15] considered edge-colourings in which only adjacent
vertices were distinguishable. After that, around 2005, Zhang et al. [14] studied the problem of distinguishable vertices in
the context of total-colourings, giving rise to AVD-total-colourings. In this article, Zhang et al. determined the AVD-total-
chromatic number for some classes of graphs and, based on their results, they posed the following conjecture:

Conjecture 1 (AVD-Total-Colouring Conjecture). If G is a simple graph, then χa′′ (G) ≤ ∆(G) + 3.

This conjecture has been verified for some classes of graphs, which include complete graphs, complete bipartite graphs,
trees [14], generalized Halin graphs with maximum degree at least six [7], graphs with ∆(G) = 3 [11,5,8], hypercubes [6],
outerplanar graphs [13], indifference graphs [10], and planar graphs with ∆(G) ≥ 14 [12]. In this work, we consider the
class of complete equipartite graphs. We prove that the AVD-total-colouring conjecture holds for this class and determine
the AVD-total-chromatic number for complete equipartite graphs of even order.

2. Preliminaries

A matching in a graph G is a set of pairwise nonadjacent edges of G. Let M be a matching of G. We say that a vertex
v ∈ V (G) is saturated by M if v is an end of an edge in M. A perfect matching is a matching that saturates all the vertices of
the graph, and a near-perfect matching is a matching in which a single vertex is not saturated.
A subset of V (G) ∪ E (G) is independent if its elements are pairwise nonadjacent and nonincident. For positive integers r
and n, a complete equipartite graph, denoted K (r × n), is a simple graph whose vertex set can be partitioned into r independent
sets (parts) of cardinality n, where any two vertices belonging to different parts are joined by an edge. In this article, we verify
the AVD-total-colouring conjecture for complete equipartite graphs. Moreover, we also determine the AVD-total-chromatic
number for even order complete equipartite graphs. We consider graphs K (r × n) with r ≥ 2 and n ≥ 2 since the results
for r < 2 or n < 2 are known [14].
A canonical labelling of K (r × n) is a labelling of the vertices of K (r × n), such that for each part j, 1 ≤ j ≤ r, each
vertex in the part receives a distinct label uij , where 1 ≤ i ≤ n, as illustrated in Fig. 1(a). For r ≥ 2, we define the canonical
decomposition [K , B ] of K (r × n) as the union of edge-disjoint subgraphs, described in the following.
Let K (r × n) be a complete equipartite graph endowed with canonical labelling. We define subgraphs K i :=
G[{ui1 , . . . , uir }], 1 ≤ i ≤ n, which are isomorphic to the complete graph Kr . Thus, K (r × n) has n disjoint copies of Kr
as induced subgraphs. Fig. 1 illustrates K (4 × 2) endowed with canonical labelling and the induced subgraphs K 1 and K 2
isomorphic to K4 .
The subgraph Bij := G[V (K i ), V (K j )] is defined as the bipartite graph with V (K i ) and V (K j ) as its parts, where 1 ≤ i <
j ≤ n. Note that Bij is a complete bipartite graph minus a perfect matching, resulting in an (r − 1)-regular graph. In fact,
j j
edges uix ux (1 ≤ x ≤ r ) do not exist since vertices uix and ux are in the same part of K (r × n). Fig. 2 illustrates a K (4 × 2)
endowed with canonical labelling and its unique bipartite subgraph, B12 , induced by the edges joining vertices from K 1 to
vertices of K 2 . Using the above notation, we define the canonical decomposition [K , B ] of K (r × n) as:
 
K := K i, and B := Bij .
1≤i≤n 1≤i<j≤n

The previous definition implies that K (r × n) ∼= (K ∪ B ). Note that, for n ≥ 2, K is a disconnected graph composed by
exactly the n components K i , 1 ≤ i ≤ n, each of which isomorphic to the complete graph Kr .
 A.G. Luiz et al. / Discrete Applied Mathematics 184 (2015) 189–195 191

(a) K (4 × 2) endowed with canonical (b) The bipartite subgraph

labelling. The bold edges induce the B12 . Note that edges u1j u2j ,
bipartite subgraph B12 . 1 ≤ j ≤ 4, do not exist.

Fig. 2. K (4 × 2) endowed with canonical labelling and its induced bipartite subgraph B12 .

(a) K (4 × 3) endowed with (b) The subgraph K composed by 3

canonical labelling. components isomorphic to K4 .

(c) A scheme showing the canonical (d) The representative graph GR

decomposition of K (4 × 3). Each square of K (4 × 3). Note that GR ∼
= K3 .
represents a component K i ⊆ K . Thick lines
joining squares represent the edges of
bipartite graphs B12 , B13 , and B23 .

Fig. 3. Canonical decomposition of K (4 × 3) and its representative graph GR .

j
Two vertices uil , up ∈ V (K (r ×n)) are called corresponding vertices if l = p and i ̸= j. That is, two vertices are corresponding
vertices if and only if they belong to the same part of K (r × n). For example, in Fig. 1, vertices u11 and u21 are corresponding
j j
vertices. Two edges uil uip , us ut ∈ E (K (r × n)) are called corresponding edges if {l, p} = {s, t }, and i ̸= j. As an example, edges
u12 u13 and u22 u23 are corresponding edges in K (4 × 2); note in Fig. 1(b) that these edges belong to distinct components K 1 and
K 2 , respectively.
Let GR be the underlying simple graph obtained from K ∪ B by shrinking each K i into a vertex vi . Graph GR is called
the representative graph of K (r × n) since the previous decomposition can be represented by GR in the following way: each
vertex vi ∈ V (GR ) represents a component K i ⊆ K and each edge vi vj ∈ E (GR ) represents a bipartite graph Bij ⊆ B . Note
that GR ∼ = Kn . For example, the representative graph of K (4 × 2) is the complete graph K2 . Fig. 3 illustrates the canonical
decomposition of K (4 × 3) and its representative graph.
In the following, we present some results that will be used in our proofs.

Proposition 2 (Zhang et al. [14]). If a simple graph G has two adjacent vertices of maximum degree, then χa′′ (G) ≥ ∆(G)+ 2. 

Theorem 3 (Zhang et al. [14]). If Kn is a complete graph of order n, n ≥ 2, then

∆(Kn ) + 2,

if n is even;
χa′′ (Kn ) =
∆(Kn ) + 3, otherwise. 

Below, we define a χa′′ (Kn )-AVD-total-colouring π and call it the standard-AVD-total-colouring of Kn .

Standard-AVD-total-colouring of Kn Let V (Kn ) := {v1 , . . . , vn }, and let k be the total of colours used by the AVD-total-
colouring π of Kn . First, we determine the value of k: if n is even, then k := ∆(Kn ) + 2; otherwise k := ∆(Kn ) + 3. Second,
192 A.G. Luiz et al. / Discrete Applied Mathematics 184 (2015) 189–195

Fig. 4. Complete graphs K5 and K6 endowed with their respective standard-AVD-total-colourings.

we assign a vertex-colouring to Kn such that, for each vi ∈ V (Kn ), π (vi ) := i. Finally, we assign an edge-colouring to Kn in
the following way: for each vi vj ∈ E (Kn ), if (i + j) is even, then π (vi vj ) := 2 , otherwise π (vi vj ) := (l)k
i +j
 i +j +k 
2 k
, where
denotes the number in {1, . . . , k} which is congruent to l (mod k).
Note that in any (∆(Kn ) + 2)-AVD-total-colouring of an even-order complete graph, there is exactly one colour that
appears only in its edges, and the edges assigned this colour form a perfect matching in Kn . Fig. 4 shows complete graphs
endowed with their respective standard-AVD-total-colourings.

Proposition 4. Let Kn be an odd order complete graph endowed with a (∆(Kn ) + 3)-standard-AVD-total-colouring. Then, for
each vertex vi with i even, C (vi ) = { 2i , n+2i+1 }; and for each vertex vi with i odd, C (vi ) = {( i−21 )n+2 , n+2i+2 }. Moreover, the set of
edges coloured with just ∆(Kn ) + 2 or with just ∆(Kn ) + 3 form near-perfect matchings.
Proof. The result follows from the definition of AVD-total-colouring. In particular, from the fact that, for odd n, there is no
vertex with colours ∆(Kn ) + 2 and ∆(Kn ) + 3. As for the second assertion, note that vn is the unique vertex where colour
∆(Kn ) + 2 does not occur, and that v1 is the unique vertex where colour ∆(Kn ) + 3 does not occur. This implies that each
one of these colours induces a near-perfect matching. 
It is well known that χ ′ (Kn ) = n, if n is odd, and χ ′ (Kn ) = n − 1, otherwise. Now, we define a particular χ ′ (Kn )-edge-
colouring λ and call it a standard-edge-colouring of Kn .
Standard-edge-colouring of Kn Let V (Kn ) := {v1 , . . . , vn }. Suppose n is odd. Then, colouring λ is defined in the following
way: for 1 ≤ i ≤ n and 1 ≤ w ≤ ⌊n/2⌋, λ(v(i+w)n v(i−w)n ) := i. Now, suppose n is even. First, colour the edges of Kn \vn as
done for n odd. Observe that colour i does not occur at vertex vi , 1 ≤ i ≤ n − 1. Then, each edge vn vi can be assigned colour i.

Theorem 5 (König [9]). If G is a bipartite graph, then χ ′ (G) = ∆(G). 

Lemma 6. Let r be an odd integer and let G[X , Y ] be an (r − 1)-regular bipartite graph with X := {x1 , . . . , xr }, Y := {y1 , . . . , yr }
and such that xi yi ̸∈ E (G), 1 ≤ i ≤ r. Then, there exists an r-edge-colouring λ: E (G) → {1, . . . , r } of G such that λ(xi yj ) ̸∈ {i, j}.
Proof. Let r be an odd integer and let H [X , Y ] ∼
= Kr ,r , where X = {x1 , . . . , xr } and Y = {y1 , . . . , yr }. Since, by Theorem 5,
χ ′ (H ) = r, let φ be an r-edge-colouring of H defined in the following way: for 1 ≤ i ≤ r and 0 ≤ p ≤ r − 1,
φ(xi y(i+p)r ) := (i + 2p)r .
Colouring φ uses r colours. Also, observe that φ(xi yi ) = i, for 1 ≤ i ≤ r. First, we prove that φ(xi yj ) ̸= φ(xi ys ) for any two
edges xi yj and xi ys . By the definition of colouring φ , j = (i + p1 )r , and s = (i + p2 )r . Since j ̸= s, we conclude that p1 ̸= p2 .
Thus, i + 2p1 and i + 2p2 lie in different congruence classes modulo r. Therefore, edges xi yj and xi ys receive distinct colours.
Now, we prove that φ(xj yt ) ̸= φ(xs yt ) for any two edges xj yt and xs yt . By the definition of colouring φ we have that
t = (j + p1 )r = (s + p2 )r , and since j ̸= s, we have p1 ̸= p2 . So, we have that (t + p1 )r and (t + p2 )r lie in different congruence
classes modulo r. Therefore, φ(xj yt ) ̸= φ(xs yt ), and φ is an r-edge-colouring of H with the property that φ(xi yi ) = i, for
1 ≤ i ≤ r.
Given the complete bipartite graph H as defined above, endowed with the r-edge-colouring φ , we use H to obtain an
r-edge-colouring λ : E (G) → {1, . . . , r } of graph G in the following way:
λ(xi yj ) := φ(xi yj ) for 1 ≤ i, j ≤ r and i ̸= j.
Since φ(xi yi ) = i, we conclude that λ(xi yj ) ̸∈ {i, j}. 

3. Results

In this section we present our main result.

Theorem 7. Let G := K (r × n) be a complete equipartite graph with r ≥ 2 and n ≥ 2. If G has even order, then χa′′ (G) =
∆(G) + 2; otherwise, χa′′ (G) ≤ ∆(G) + 3.
Proof. Let G := K (r × n), r ≥ 2 and n ≥ 2, endowed with canonical labelling, and [K , B ] be a canonical decomposition of
G. By Proposition 2, χa′′ (G) ≥ ∆(G) + 2 = n(r − 1) + 2.
 A.G. Luiz et al. / Discrete Applied Mathematics 184 (2015) 189–195 193

We consider four cases depending on the parity of n and r. For each case, we construct an AVD-total-colouring π for G
with ∆(G) + 2 colours, if G has even order, or with ∆(G) + 3 colours, if G has odd order. In order to do this, we construct an
AVD-total-colouring φ for K and an edge-colouring λ for B . Colouring π is defined in such way that φ is a restriction of π
to the elements of K , and λ a restriction of π to the edges of B .

Case 1. r , n ≡ 0 (mod 2).

In this case, we first construct φ using r + 1 colours, and then λ with (n − 1)(r − 1) additional colours. Therefore, π uses
n(r − 1) + 2 = ∆(G) + 2 colours.
By Theorem 3, χa′′ (K i ) = r + 1. For each i, 1 ≤ i ≤ n, adjust notation so that V (K i ) = {ui1 , . . . , uir }. Let φ i be a standard-
AVD-total-colouring of K i that uses r + 1 colours such that for any φ i and φ j , i ̸= j, corresponding vertices and edges of G
receive the same colour.
Thus, colouring φ is defined as follows.

φ(uij ) := φ i (uij ), uij ∈ V (K i ), 1 ≤ i ≤ n and 1 ≤ j ≤ r ;

φ(uij uik ) := φ i (uij uik ), uij uik ∈ E (K i ), 1 ≤ i ≤ n and 1 ≤ j, k ≤ r .

Note that the AVD-total-colourings φ i determine an AVD-total-colouring φ to K since every two adjacent vertices of K
belong to the same component.
Now, it remains to colour the edges of B . Each Bij ⊆ B is an (r − 1)-regular bipartite graph that will be coloured with
r − 1 colours. In order to obtain a way of reusing the same set of colours for disjoint bipartite graphs Bij , we consider the
colour classes of a χ ′ (GR )-edge-colouring of representative graph GR . Note that in such a colouring, each colour class is a
perfect matching of GR . Also, each edge vi vj of GR is assigned to the bipartite graph Bij . Therefore, there exists a set F of
disjoint bipartite graphs Bij assigned to each perfect matching of GR . We conclude that, to edge-colour the elements of each
set F , a total of r − 1 colours are required. Since χ ′ (GR ) = n − 1, there exist exactly n − 1 sets F , which implies (n − 1)(r − 1)
new colours used to colour the edges of B . Let λ be the edge-colouring just constructed for B .
We claim that π is a total-colouring. Note that every adjacent or incident elements of K have distinct colours because K
was assigned an AVD-total-colouring. Moreover, every two adjacent edges of B have distinct colours because B received
an edge-colouring. Since all colours used in the AVD-total-colouring of K are distinct from the colours used in the edge-
colouring of B , every two incident or adjacent elements of G have distinct colours. Therefore, π is a total-colouring of G.
For a vertex uil ∈ V (G), note that Cπ (uil ) = Cφ (uil ) ∪ Cλ (uil ). In order to conclude this case, we have to prove that
j j j
Cπ (uil ) ̸= Cπ (up ) for any two adjacent uil , up ∈ V (G). If j = i, vertices uil , up belong to the same component of K , and
j j
the result follows by construction of φ . Suppose that i ̸= j. By construction of φ , Cφ (uil ) ̸= Cφ (up ) because uil , up are not
j
corresponding vertices. Moreover, all the colours used to edge-colouring B occur in both and up ; uil ( ) = Cλ (ujp ).
that is, Cλ uil
j
Nevertheless, the set of colours used by φ λ i
is disjoint from that used by . We conclude that Cπ ul ( ) ̸= ( )
Cπ up .

Case 2. r ≡ 1 (mod 2) and n ≡ 0 (mod 2).

In this case, we first construct φ using 2r colours, and then λ with (n − 2)(r − 1) + r colours, where r of them are used
in φ . Therefore, π uses n(r − 1) + 2 = ∆(G) + 2 colours.
For each i, 1 ≤ i ≤ n, adjust notation so that V (K i ) = {ui1 , . . . , uir }. Since r is odd, χ ′ (K i ) = r. We construct an AVD-total-
colouring φ i for K i in the following way: (i) the edges of K i receive a standard-r-edge-colouring with colours from the set
{1, . . . , r }; (ii) φ i (uij ) := r + j, 1 ≤ j ≤ r. According to the definition of φ i , for each pair of adjacent vertices uij , uil ∈ V (K i ),
we have C φ i (uij ) = {j} ∪ {r + x: 1 ≤ x ≤ r and x ̸= j} which is different from C φ i (uil ) = {l} ∪ {r + x: 1 ≤ x ≤ r and x ̸= l}.
Thus, uij and uil have distinct sets of colours. Now, colouring φ is defined as in the previous case and we conclude that φ is a
(2r )-AVD-total-colouring by the same reasoning.
Now it remains to colour the edges of B . As we already know, it is possible to colour the edges of each bipartite Bij with
r − 1 colours. Graph GR has n − 1 disjoint perfect matchings. For n − 2 of them, we apply the same reasoning of Case 1,
using (n − 2)(r − 1) new colours. Let M be the remaining perfect matching of GR . For each Bij assigned to an edge of M, let
λij : E (Bij ) → {r + 1, . . . , 2r } be an r-edge-colouring obtained from Lemma 6, such that λij (uil ujp ) ̸∈ {r + l, r + p}. Therefore,
the edges of B received (n − 2)(r − 1) + r colours with r of them also used in the vertices of K . Let λ be the edge-colouring
just constructed for B .
We claim that π is a total-colouring for G. By construction, φ is an AVD-total-colouring for K . Also, by construction, λ
is an edge-colouring for B . For edge-colouring B we use (n − 2)(r − 1) new colours and reuse the r colours assigned to
vertices of G. Recall that colours used in the vertices of G are not assigned to any edge of K . By construction of φ and λ, and by
Lemma 6, we conclude that there are no edges of B incident with a vertex of same colour. Therefore, π is a total-colouring.
Colouring π is also adjacent-vertex-distinguishing. Note that any two adjacent vertices belonging to the same compo-
nent K i have distinct sets of colours, which is guaranteed by the AVD-total-colouring given to K . Since the colours used to
edge-colour B occur on each vertex of G, any two adjacent vertices u and v belonging to different components of K are
also distinguished by the AVD-total-colouring given to K . Thus, C (u) ̸= C (v) for any two adjacent vertices u, v ∈ V (G).
Therefore, π is a (∆(G) + 2)-AVD-total-colouring of G.
194 A.G. Luiz et al. / Discrete Applied Mathematics 184 (2015) 189–195

Case 3. r ≡ 0 (mod 2) and n ≡ 1 (mod 2).

In this case, we first construct λ using n(r − 1) colours, and then φ with r + 1 colours, using just two new colours.
Therefore, π uses n(r − 1) + 2 = ∆(G) + 2 colours.
Since n is odd, GR is Class 2. Consider a standard-edge-colouring for GR . By the same reasoning of previous cases, we
obtain an n(r − 1)-edge-colouring λ for B from this edge-colouring of GR . By construction of the standard-edge-colouring
of GR , colour i does not occur in vi ∈ V (GR ), for each 1 ≤ i ≤ n. This implies that there exists a set of r − 1 colours of λ that
do not occur in vertices of K i . Let Si be this set of colours and define Si := {c1i , . . . , cri −1 }.
Now, we construct φ , an (r + 1)-AVD-total-colouring for K . In order to do this, we define φ i for each K i using colours
Si ∪ {x, y}, 1 ≤ i ≤ n. Recall that in any (∆(Kn ) + 2)-AVD-total-colouring of an even-order complete graph, there is exactly
one colour which induces a perfect matching. For each K i , adjust notation so that: (i) colour y induces this perfect matching;
(ii) φ i (uir ) = x; and (iii) φ i (uij ) = cji , 1 ≤ j ≤ r − 1. Note that vertices assigned colour x are non-adjacent, since they are in
the same part of G.
In order to conclude this case, we prove that π , as previously defined, is a (∆(G) + 2)-AVD-total-colouring of G. By
construction of λ and φ , colouring π uses n(r − 1) + 2 = ∆(G) + 2 colours.
We claim that π is a total-colouring. To see this, observe that the AVD-total-colouring assigned to K guarantees that any
two adjacent or incident elements in the same component K i receive distinct colours. Moreover, any two adjacent edges of
j
subgraph B receive distinct colours because λ is an edge-colouring. In order to see that π (uik ) ̸= π (uik ul ), i ̸= j, remember
the definition of Si . Colours of Si do not occur in any edge of B , 1 ≤ p ≤ n, p ̸= i. Moreover, π (uk ) ∈ Si ∪ {x}, and x is not
ip i

j
used to colour edges of B . It remains to analyse two adjacent vertices of G, uik and ul , i ̸= j. Colour π (uik ) ∈ Si ∪ {x} and colour
j j
π( ) ∈ Sj ∪ {x}. By construction, Si and Sj are disjoint and at most one of { , } received colour x since these vertices are
ul uik ul
adjacent. Therefore, we conclude that π is a total-colouring.
Now we prove that π is also adjacent-vertex-distinguishing. Since φ i , 1 ≤ i ≤ n, is an AVD-total-colouring, we have just
j
to analyse adjacent vertices uil and uk , i ̸= j, in distinct components of K . For each Si ∪ {x}, 1 ≤ i ≤ n, define (c1i , . . . , cri −1 , x)
j
as an order for this set. Consider the corresponding vertices uik and uk of G. By the definition of π , C π (uik ) = c, c ∈ Si ∪{x}, and
j
C π (uk ) = c , c ∈ Sj ∪ {x}. Moreover, c and c have the same position in the previously defined order of Si ∪ {x} and Sj ∪ {x}.
′ ′ ′
j
By the definition of φ i , for each uil , l ̸= k, C π (uil ) ̸= C π (uik ). Since Si ∩ Sj = ∅ and k ̸= l, we conclude that C π (uil ) ̸= C π (uk ).

Case 4. r , n ≡ 1 (mod 2).

This is the only case for which G has odd order. In this case, we construct λ using n(r − 1) colours, and φ with r + 2
colours, using just three new colours. Therefore, π uses n(r − 1) + 3 = ∆(G) + 3 colours.
Like Case 3, here n is odd and GR is Class 2. We construct a standard edge-colouring λ for B that uses n(r − 1) colours.
We define Si := {c1i , . . . , cri −1 } as the set of colours of λ that do not occur in K i , as in the previous case.
In this case, each K i is an odd-order complete graph. So, we construct an (r + 2)-AVD-total-colouring φ for K . In order
to do this, we define a standard-AVD-total-colouring φ i for each K i using colours Si ∪ {x, y, z }, 1 ≤ i ≤ n. Recall that, in
a standard-AVD-total-colouring of an odd-order complete graph, each of the two colours that appear only in edges of Kn
induces a near-perfect matching. For each K i , adjust notation so that φ i (uij ) = cji , 1 ≤ j ≤ r − 1 and φ i (uir ) = x. Colours y
and z appear only in edges of K i . Note that vertices assigned colour x are non-adjacent since they are in the same part of G.
In order to conclude this case, we prove that π , as previously defined, is a (∆(G) + 3)-AVD-total-colouring of G. By con-
struction of λ and φ , colouring π uses n(r − 1) + 3 = ∆(G) + 3 colours. Colouring π is an AVD-total-colouring by exactly
the same reasoning of the previous case. 
In Theorem 7, the AVD-total-colouring of complete equipartite graphs of even order is determined. For graphs of odd
order in this family, we only verify the AVD-Total-Colouring Conjecture. Moreover, we did not find complete equipartite
graphs of odd order that needed ∆(K (r × n)) + 3 colours. Based on these observations, we pose the following conjecture:

Conjecture 8. If K (r × n) has odd order, then χa′′ (K (r × n)) = ∆(K (r × n)) + 2.

References

[1] P. Balister, B. Bollobás, R. Schelp, Vertex distinguishing colorings of graphs with ∆(G) = 2, Discrete Math. 252 (2) (2002) 17–29. http://dx.doi.org/10.
1016/S0012-365X(01)00287-4.
[2] C. Bazgan, A. Harkat-Benhamdine, H. Li, M. Woźniak, On the vertex-distinguishing proper edge-coloring of graphs, J. Combin. Theory Ser. B 75 (2)
(1999) 288–301. http://dx.doi.org/10.1006/jctb.1998.1884.
[3] A.C. Burris, R.H. Schelp, Vertex-distinguishing proper edge colorings, J. Graph Theory 26 (2) (1997) 73–82.
[4] G. Chartrand, M.S. Jacobson, J. Lehel, O.R. Oellerman, S. Ruiz, F. Saba, Irregular networks, Congr. Numer. 64 (1988) 197–210.
[5] X. Chen, On the adjacent vertex distinguishing total coloring numbers of graphs with ∆(G) = 3, Discrete Math. 308 (2008) 4003–4007. http://dx.doi.
org/10.1016/j.disc.2007.07.091.
[6] M. Chen, X. Guo, Adjacent vertex-distinguishing edge and total chromatic numbers of hypercubes, Inform. Process. Lett. 109 (12) (2009) 599–602.
http://dx.doi.org/10.1016/j.ipl.2009.02.006.
[7] X. Chen, Z. Zhang, AVDTC numbers of generalized Halin graphs with maximum degree at least 6, Acta Math. Appl. Sin. Engl. Ser. 24 (1) (2008) 55–58.
http://dx.doi.org/10.1007/s10255-005-5222-8.
 A.G. Luiz et al. / Discrete Applied Mathematics 184 (2015) 189–195 195

[8] J. Hulgan, Concise proofs for adjacent vertex-distinguishing total colourings, Discrete Math. 309 (2009) 2548–2550. http://dx.doi.org/10.1016/j.disc.
2008.06.002.
[9] D. König, Über Graphen und ihre Anwendung auf Determinantentheorie und Mengenlehre, Math. Ann. 77 (1916) 453–465.
[10] V. Pedrotti, C.P. de Mello, Adjacent-vertex-distinguishing total colouring of indifference graphs, Mat. Contemp. 39 (2010) 101–110.
[11] H. Wang, On the adjacent vertex-distinguishing total chromatic numbers of the graphs with ∆(G) = 3, J. Comb. Optim. 14 (1) (2007) 87–109.
http://dx.doi.org/10.1007/s10878-006-9038-0.
[12] W. Wang, D. Huang, The adjacent vertex distinguishing total coloring of planar graphs, Journal of Combinatorial Optimization 27 (2) (2014) 379–396.
http://dx.doi.org/10.1007/s10878-012-9527-2.
[13] Y. Wang, W. Wang, Adjacent vertex distinguishing total colorings of outerplanar graphs, Journal of Combinatorial Optimization 19 (2) (2010) 123–133.
http://dx.doi.org/10.1007/s10878-008-9165-x.
[14] Z. Zhang, X. Chen, J. Li, B. Yao, X. Lu, J. Wang, On adjacent-vertex-distinguishing total coloring of graphs, Sci. China Ser. A: Math. 48 (3) (2005) 289–299.
http://dx.doi.org/10.1360/03ys0207.
[15] Z. Zhang, L. Liu, J. Wang, Adjacent strong edge coloring of graphs, Appl. Math. Lett. 15 (5) (2002) 623–626. http://dx.doi.org/10.1016/S0893-9659(02)
80015-5.