On the equality of the grundy numbers of a graph

International Journal of Next-Generation Networks (IJNGN) Vol.5, No.4, December 2013

On The Equality Of The Grundy Numbers Of A
Graph
Ali Mansouri and Mohamed Salim Bouhlel
Department of Electronic Technologies of Information and Telecommunications Sfax,
Tunisia

ABSTRACT
Our work becomes integrated into the general problem of the stability of the network ad hoc. Some, works
attacked(affected) this problem. Among these works, we find the modelling of the network ad hoc in the
form of a graph. Thus the problem of stability of the network ad hoc which corresponds to a problem of
allocation of frequency amounts to a problem of allocation of colors in the vertex of graph. we present use
a parameter of coloring " the number of Grundy”. The Grundy number of a graph G, denoted by Γ(G), is
the largest k such that G has a greedy k-coloring, that is a coloring with colours obtained by applying the
greedy algorithm according to some ordering of the vertices of G. In this paper, we study the Grundy
number of the lexicographic, Cartesian and direct products of two graphs in terms of the Grundy numbers
of these graphs.

KEYWORDS
colouring, greedy algorithm, graph product ,grundy number, on-line algorithm,

1. INTRODUCTION
Graphs considered in this paper are undirected, finite and contain neither loops nor multiple
edges (unless stated otherwise). The definitions and notations used in this paper are standard and
may be found in any textbook on graph theory. A (proper) k-colouring of a graph G=(V,E) is a
mapping c:V →{1,...,k}, such that for any edge uv ∈ E(G),c(u) ≠c(v).
A k-colouring may also be seen as a partition of the vertex set of G into k disjoint stable set.
Si={v |c(v)=i} for 1≤i≤k. For convenience , by k-colouring we mean either the mapping c or the
partition(S1,...,Sk).The elements of {1,...,k} are called colours. A graph is k-colourable if it has a
k-colouring. The chromatic number χ(G) is the least k such that G is k-colourable. Several online
algorithms producing colourings have been designed.
The most basic and most widespread one is the greedy algorithm. A greedy colouring relative
to a vertex ordering σ=v1<v2<···<vn of V(G) is obtained by colouring the vertices in the order
v1,...,vn, assigning to vi the smallest positive integer not already used on its lowered-indexed
neighbours. Denoting by Si the stable set of vertices coloured i, a greedy colouring has the
following property:
For every j<i, every vertex in Si has a neighbour in Sj (1)
Otherwise the vertex in S would have been coloured j. Conversely, a colouring satisfying
―Property (1)‖ is a greedy colouring relative to any vertex ordering in which the vertices of Si
precede those of Sj when i < j. The Grundy number Γ(G) is the largest k such that G has a greedy
k-colouring. It is well known that
DOI : 10.5121/ijngn.2013.5402

17


ω(G) ≤χ(G)≤Γ(G)≤Δ(G)+1 where ω(G) denotes the clique number of G and Δ(G)the maximum
degree of G.
The inequality χ(G)≤Γ(G) may be tight, but it can also be very loose. Zaker [15] showed that for
any fixed k≥ 0, given a graph G it is CoNP-Complete to decide whether Γ(G)≤χ(G)+k. He also
showed that, given a graph G which is the complement of bipartite graph, it is CoNP-Complete to
decide if Γ(G)=χ(G). This implies that it is CoNP-Complete to decide if Γ(G)=ω(G). Indeed, if G
is the complement of a bipartite graph, then it is perfect, so χ(G)=ω(G).
The Grundy number of various classes of graphs has been studied (see the introduction of [1]).
In this paper, we study the grundy number of different usual products of two graphs G and H.
The lexicographic product G[H],the direct product G×H, and the Cartesian product G □ H, of G
by H are the graphs with vertex set V(G) ×V(H)and the following edge set:
E(G[H]) = {(a,x) (b,y ) | ab ∈ E(G), or a=b and xy ∈ E(H)};
E(G×H) = {(a,x) (b,y) | ab ∈ E(G), and xy ∈ E (H)};
E(G□H) ={(a,x) (b,y) | a=b and xy ∈ E(H) or ab ∈ E(G) and x=y}.
It follows from the definition that G ×H (resp. G□H ) and H×G (resp. H□G ) are isomorphic.
But G[H] and H[G] are generally not isomorphic. Moreover G[H] may be seen as the graph
obtained by blowing up each vertex of G into a copy of H.
Regarding the lexicographic product, we prove in Section 3 that for any graphs G and H
Γ(G)×Γ(H)≤Γ(G[H])≤2 Γ(G)−1(Γ(H)−1)+Γ(G)−1
In addition, we show that if G is a tree or Γ(G)=Δ(G)+1, then Γ(G[H])=Γ(G)×Γ(H). Using these
results, we prove a stronger complexity result than the one of Zaker [15] mentioned above: for
every fixed c ≥1,it is CoNP-Complete to decide if Γ(G)≤ c×χ (G) for a given graph G.
Analogously, we show that it is CoNP-Complete to decide if Γ(G) ≤c×ω(G).
In Section 4, we investigate the Grundy number of the cartesian product of two graphs. We show
that (G□H) ≥max{Γ(G),Γ(H)} and increase this lower bound in some particular cases. We prove
that there is no upper bound of Γ(G□H) as a function of Γ(G) and Γ(H). More precisely we show
that for the complete bipartite K p,p, Γ(K p,p)=2 but Γ (K p,p □ K p,p) ≥ p+1.
Nevertheless, we show that for any fixed graph G, there is a function hG such that, for any graph
H, Γ(G□H) ≤hG(Γ(H)).
in fact, we show that hG(k) ≤Δ(G)·2 k−1+k
We then give a better upper bound for hG (2) for some graphs G.
Finally, in Section 5, we study the Grundy number of the direct product of two graphs. We show
that Γ (G×H)≥Γ(G)+Γ(H)−2 and construct for any k some graph Gk such that Γ(Gk)=2k+1 and Γ
(Gk×K2)=3k+1.

2. PRELIMINAIRES
In this section, we present some definitions and preliminary results.
A sub graph of a graph G is a graph H such that V (H) ⊂V (G) and E(H)⊂E(G). Note that since
18


H is a graph we have E (H) ⊂ E (G) ∩ [V (H)] 2. If H contains all the edges of G between vertices
of V (H), that is E (H) =E (G) ∩V (H)] 2, then H is the subgraph induced by V (H).
If S is a set of vertices, we denote by G<S> the graph induced by S and by G−S the graph induced
by V (G)S.
For simplicity, we write G −v rather than G−{v}. For a subset F of E(G), we write GF
= (V (G), E (G)F).
For a subset F of E (G), we write GF = (V (G), E (G)F). As above G {e}is abbreviated to Ge.
If H is a subgraph of G then χ(H)≤χ(G). This assertion cannot be transposed to the grundy
number. For example, the path P4 of order 4 is a subgraph of the cycle C4 of order 4 but one can
easily check that Γ(P4)=3 and Γ(C4)=2. However such an assertion holds if we add the extra
condition of being an induced subgraph.
Proposition1 If H is an induced subgraph of G then Γ(H)≤Γ(G).
Proof.
Let σ be an ordering for which the corresponding greedy colouring of H uses Γ (H) colours.
Then a colouring with respect to any ordering of V (G) beginning with σ will use at least Γ (H) to
colour H, hence at least Γ (H) to colour G.
Lemma2 Let G be a graph and u and v two vertices G. The following hold:
For any edge e, Γ(G) −1≤Γ(Ge)≤Γ(G)+1. (1)
If N (u) ⊂ N (v) then in every greedy colouring c of G, c(u)≤c(v). In particular, if N (u) =N (v)
then:
c (u)=c(v). (2)
If N (u) =N (v) then Γ (G) =Γ (G−u). (3)
Proof.
(1) Set e= xy and p= Γ (G). Let (S1,...,Sp) be a greedy p-colouring of G. It satisfies Property (1).
Let i be the integer such that x ∈ Si and let Tj=Sj for1≤ j< i and Tj=S j+1 for i≤ j≤ p−1.
It is a simple matter to check that (T1,...,T p −1) satisfies Property (1). Hence Γ (G−Si) ≥ p-1. As
G−Si is an induced subgraph of Ge, by proposition1, Γ (Ge) ≥ p−1.
Set q=Γ (G). Let (S1’,...,Sq’) be a greedy q-colouring of Ge. It satisfies Property (1). Now let i be
the integer such that x ∈ Si’. Let Tj’= Sj’ for 1≤ j<i and Tj’= S j+1’ for i≤ j≤q−1. It is a simple matter
to check that (T1’,..., T q-1’) satisfies Property (1). Hence Γ(G−Si)≥ q−1. As G−Si is an induced
subgraph of G, by Proposition 1, Γ (G) ≥q−1.
(2) Let c= (S1,...,Sp). Suppose u ∈Sj and v ∈ S i. Since v ∈ Si, then v has no neighbour in Si. So u
has no neighbour in Sj because N (u) ⊂N (v). Thus j≤ i because c satisfies Property (1).
(3) Let S1,...,Sp be the stable sets of a greedy couloring..
By (2), u and v are in the same stable set Si.
Now S1,...,S i−1,Si{u},S i+1,...,Sp are the stable sets of a greedy colouring of G−u. Indeed as
NG (u) =NG (v) it is a simple matter to check that they satisfy Property (1).

19


A path is a non-empty graph P= (V, E) of the form V= {x0,x1,...,xk} and E={x0x1,x1x2,...,x k−1 xk}
where the xi are all distinct. The vertices x0 and xk are the endvertices of P. A (u,v)-path is a path
with endvertices u and v. A graph is connected if for any two vertices u an v there is a (u,v)- path.
Proposition3 Let G be a connected graph. Then Γ(G)=2 if and only if G is complete bipartite.
Proof. It is easy to see that if G is complete bipartite then Γ(G)=2: indeed applying several times.
Lemma 2 (3), we obtain that Γ (G) =Γ (K2) =2.
Conversely, if Γ (G) =2, then G has to be bipartite because Γ(G)≥χ(G). Suppose now that G is not
complete bipartite. Then there exist two vertices u and v in different parts of the partition which
are not adjacent. Let P be a shortest (u,v)-path. Then P has odd length, so length at least 3 and
because it is a shortest path it is an induced path. Hence G contains an induced P4. So by
Proposition 1, Γ (G) ≥3.
This proposition implies that one can decides in polynomial time if the Grundy number of a graph
is 2. More generally, Zaker [15] showed that for any fixed k, it is decidable in polynomial time if
a given graph has grundy number at most k. To show this, he proved that there is a finite number
of graphs called k-atoms such that if Γ(G) ≥ k then G contains a k-atom as induced
subgraph. The k-atoms may easily be found using Proposition 5 below.
Definition4 Let G be a graph and W a subset of V (G). A set S is W-dominating if S ⊂V (G)W
and every vertex of W has a neighbour in S.
The following proposition follows immediately from the Property (1) of greedy colouring.
Proposition5 Let G be a graph and W a subset of V (G). If S is a W-dominating stable set then
Γ (G<W∪S>) ≥Γ (G<W>)+1.
Note that if S is a W-dominating set then Γ(G<W∪S>) cannot be bounded by a function of
Γ (G<W>).
For example, a tree may be partitioned into two stable sets S and T. Moreover, because the tree is
connected S is T-dominating (and vice-versa). But the Grundy number of a stable set is 1 whereas
the Grundy number of a tree may be arbitrarily large. Consider for example the binomial tree of
index k Tk which may be defined recursively as follows:
• T1 is the graph with one vertex and no edge;
• Tk is constructed from T k −1 by joining each vertex to a new leaf.
The binomial tree Tk has chromatic number 2 and grundy number k. It is the unique k-atom which
is a tree. Hence, as shown in [8], the Grundy number of a tree is the largest index of a binomial
tree it contains.
The union of two graphs G1 and G2 is the graph G1 ∪G2 with vertex set V (G1) ∪V (G2) and edge
set E(G1) ∪E(G2). If G1 and G2 are disjoint (i.e. V (G1) ∩V (G2) = ϕ), we refer to their union as a
disjoint union and denote it G1+G2. The join of two disjoint graphs G1 and G2 is the graph G1 ⊕G2
obtained from G1+G2 by joining all the vertices of G1 to all the vertices of G2.
Proposition6

If G=G1+G2 then Γ (G) =max (Γ (G1), Γ (G2)).
If G = G1⊕G2 then Γ (G) =Γ (G1) +Γ (G2).

This proposition and an immediate induction yield a result of Gyarf´as and Lehel [6] stating that
for every cograph (graph without induced P4) Γ (G) =χ (G) because every cograph of order at
least two is either the disjoint union or the join of two cographs.

20


Lemma7 Let G be a graph and x a vertex of G. If there is a greedy colouring c such that x is
coloured p then for any 1≤i≤ p, there is a greedy colouring such that x is coloured i.
Proof. For 1≤i≤ p−1, let Si be the stable set of vertices coloured i by c. Then for any 1≤i≤
p,(S1,...,S i−1,{x}) is a greedy i-colouring of G<{x}∪∪ i−1 j=1 Si> in which x is coloured i. This
partial greedy colouring of G may be extended into a greedy colouring of G in which x is
coloured i.
Lemma8 Let G be a graph with at least one edge. There are two adjacent vertices x and y such
that there is two greedy colourings cx and cy such that cx(x) =cy(y) = Γ (G).
Proof. Set p= Γ (G) and let cx be a greedy p-colouring of G with stable sets S1,...,Sp. Let x a vertex
of Sp and y a neighbor of x in S p−1. Then S1,S2,...S p−2,Sp,{y}is a partial greedy colouring cy of G
with cy(x)= p−1 and cy(y)= p. This colouring may trivially be extended to G.

3. LEXICOGRAPHIC PRODUCT
Obviously, χ(G[H])≤χ(G)×χ(H) and Stahl [11] showed χ(G[H])≥χ(G)+2χ(H)−2. In this section,
we establish some bounds on Γ(G[H]) in terms of Γ(G) and Γ(H).
Definition9 [12] In the lexicographic product G[H], for every vertex x∈ G, we call copy of H at x
the graph H(x) isomorphic to H which is induced by the vertices of {x}×V(H).
Proposition10 Let G and H be two graphs. In a greedy colouring of G[H], at most Γ(H) colours
appear on each H(x), x ∈ V(G).
Proof. Consider a greedy colouring of G[H] and let n1,n2,...,np be the p colours appearing on a
particular copy H(x) of H. For any 1≤i≤ p, let Si be the stable set of vertices of H(x) coloured ni.
Let u be a vertex of Si. For any 1≤ j<i, by the Property (1), in G[H], u has a neighbour v coloured
nj. The vertex v must be in H(x) because the neighbours of x not in H(x) are also neighbours of
the vertex z of H(x) coloured nj. Hence v∈ Sj. It follows that the colouring (S1,...,Sp) satisfies the
Property (1). Hence Γ(H)= Γ(H(x))≥ p. Geller and Stahl [5] showed that if χ(H)=k then
χ(G[H])=χ(G[Kk]) for any graph G. We now prove a similar result for the Grundy number.
Theorem11 Let H be a graph such that Γ(H)=k. Then for any graph G, Γ (G [H]) =Γ (G [Kk]).
Proof. Set V (G) = {v1,...,vn}.
Let c be a greedy colouring of G [H]. For every 1≤i≤n, let Ai =c(H(Vi))={α1 i,...,α|Ai| i } be the set of
colours appearing on H(vi). Let F be the graph obtained from G[H] by replacing each H(vi) by a
complete graph on |Ai| vertices,w1 i,...,w|Ai| i and c’ be the colouring of F defined by c’(wj i)= αj i for
any 1≤i≤n and 1≤ j ≤|Ai|.
By construction F is an induced subgraph of G [Kk] because for each1≤i≤n, |Ai|≤k by
Proposition10. Moreover, it is a simple matter to check that c’ is a greedy colouring of F. Hence Γ
(G [Kk]) ≥Γ (F) ≥Γ (G [H]).

3.1. Lower bounds
Proposition 12 Let G and H be two graphs. Then Γ(G[H]) ≥Γ(G)×Γ(H).
Proof of Proposition 12. Let cG (resp. cH) be a greedy colouring of G (resp. H) with Γ (H) (resp.Γ
(G)) colours. Then the colouring c=(cG,cH) with the pairs of colours ordered according to the
lexicographic product is a greedy colouring of G[H].

21


Proposition 12 is tight as there are pairs of graphs (G, H) for which Γ (G [H]) =Γ (G) ×Γ (H). In
particular, we shall prove that if G is a tree or satisfies Γ(G)=Δ(G)+1 this is the case.
Theorem 13 Let G be and H be two graphs. If Γ(G)=Δ(G)+1 then Γ(G[H])=Γ(G)×Γ(H).
Proof. By Proposition 12, Γ (G [H]) ≥Γ (G) ×Γ (H).
Let us now show that Γ (G [H]) ≤Γ (G) ×Γ (H). Consider a greedy colouring of G [H]. Let u be a
vertex of G [H] coloured with the largest colour cmax and H(x) the copy of H containing u.
Moreover by the the Property (1), every colour but cmax must appear on the neighbour hood of u.
Hence cmax ≤Γ (G) ×Γ (H).
Theorem 14 Let T be a tree and H be an arbitrary graph. Then Γ(T[H])=Γ(T)×Γ(H).
Proof. Let k be the integer such that k Γ(H)≥Γ(T[H])≥(k−1)Γ(H)+1. We will prove that Γ (T) ≥k
by showing that T contains a binomial tree of index k as an induced subgraph. This implies that Γ
(T [H]) ≤Γ (T) ×Γ (H).
So by Proposition 12, Γ (T [H]) =Γ (T) ×Γ (H).
Let f be a greedy colouring of T[H] with Γ(T[H]) colours in the following, by colour we should
understand colour assigned by f. We shall construct step by step a binomial tree of order k in T.
Step 1: Let v1 be a vertex of T such that a vertex of H(v1) is coloured c1=Γ(T[H]). Then the sub
tree of T with unique vertex v1 isT1. Let P1(v1) be the sequence (v1).
Step i: (for 2≤i≤k) we have the binomial sub tree T i−1 of T. Moreover, to each vertex v of T
associated a sequence P i−1(v)=(v i−1,v i−2,...,v2,v1) of i−1 vertices in T i−1 such that :

i−1

is

(a) P i−1 (v) contains v and all its neighbours in T i−1, and
(b) for all 1≤ j≤i−1, H(vj)contains the greatest colour not appearing on∪ j-1 l=1 H(vl).
We shall construct Ti, that is add a leaf to each vertex of T i −1, and also describe the sequences Pi
satisfying the conditions (a) and (b). Let v be a vertex of T i−1.
As Pi−1 (v) contains i−1 vertices, at most (i −1) Γ (H) colours appear on ∪i-1 l=1H(vl)
byProposition10.

3.2. Upper bounds
There are pairs of graphs (G, H) for which Γ (G [H]) is greater than Γ (G) ×Γ (H) as shown by the
following proposition.
Proposition 15 Let G3 be the graph depicted. Then Γ(G3)=3 and Γ(G3[K2p]) ≥7p

22


Proof. Let us first show that Γ (G3) =3.
Suppose, by way of contradiction, that G3 admits a greedy 4-colouring. Then one of the two
vertices of degree three, namely a and b, is coloured 4. By symmetry, we may assume that it is a.
This vertex must have a neighbour coloured 3. This neighour is necessarily b which is the unique
one having degree at least two in G 3 −a. The vertices a and b must each have a neighbour
coloured 2 which must have degree at least one in G− {a,b}. Hence f and c are coloured 2. These
two vertices must have a neighbour coloured 1. So d and e are coloured 1, which is a
contradiction as they are adjacent.
[

Let us now show that Γ(G3[K2p]) ≥7p. For every vertex v ∈V (G3), let us assign 2p colours to the
2p vertices of K2(v) as follows. I3∪I5 to K2(a), I6∪I7 to K2(b), I2∪I4 to K2(c), I1∪I3 to K2(d), I2∪I5 to
K2(e), I1∪I4 to K2(f), I1∪I2 to K2(g) and K2(h). It is a simple matter to check that this is a greedy 7pcolouring of G3[K2p]. We would like to find upper bounds on Γ (G [H]) in terms of
Γ(G) and Γ(H).
Ideally we would like to determine exactly
Ψ (k,l) = max {Γ(G[H])| Γ(G)=k and Γ(H)=l}
= max {Γ (G [Kl])| Γ (G) =k}
by Theorem 11. In the remainder of this section we give upper and lower bounds on ψ. Note that
Γ (G) =1 if and only if G has no edge. Thus if Γ (H) =1 then Γ (G [H]) =Γ (G) using Lemma 2(3).Moreover if Γ (G)= 1 then Γ(G[H])=Γ(H) by Proposition 6. In the remainder of the section,
we will assume that all the graphs we consider have Grundy number at least 2.
Theorem 16 Γ(G[H])≤2 Γ(G)−1(Γ(H)−1)+Γ(G)−1
Proof. Let k be the integer such that 2 k−1(Γ(H)−1)+k−1≥Γ(G[H])> 2 k−2(Γ(H)−1)+k−2.
We will show that Γ(G)≥k, which implies that Γ(G[H])≤2 Γ(G)−1(Γ(H)−1)+Γ(G)−1.
Let f be a greedy colouring of G [H] with Γ(G[H]) colours. We shall construct step by step an
induced subgraph of G which has Grundy number at least k.
Step 1: Let v1 be a vertex such that the largest colour c1 =Γ (G[H]) appears on H(v1). Let
G1=G< {v1}>. Then Γ (G1) =1.
Step 2: Since Γ(G[H])>2 k−2(Γ(H)−1)+k−2≥Γ(H), by Proposition10,there are colours that do not
appear on H(v1). Let c2 be the largest such colour. For c1 >c2, there is a vertex v2∈ NG(v1) such
that c2 appears on H(v2). Let G2=G <{v1,v2}>. Since v1v2 in an edge, Γ(G2)=2
Step i: (for 3≤i≤k): We have a subgraph G i−1 of G of at most 2 i−2 vertices such that Γ (G i−1) ≥ i
−1and at most 2 i−2(Γ(H)−1)+i−2colours appear on G i−1[H]. For Γ(G[H]) > 2 k−2 (Γ(H)−1)+ k−2≥2
i−2 (Γ(H)−1) + i−2,there are colours that do not appear on G i−1[H]. Let ci be the greatest such
colour. Since c0 >c1 >...>ci and ci does not appear on G i−1[H], any vertex v ∈V(Gi) has a neighbour
n(v) in V(G)V(Gi) such that the colour ci appears on H(n(v)). Let Si= {n (v), v ∈ V (Gi)} and Gi =
G<V(Gi−1) ∪Si>. Then |Si|≤|G i−1| so |Gi|≤ 2 i−1. Moreover Si is a stable set since the colour ci
appears on the copy of H at each vertex of Si. So by Proposition5, Γ (Gi) ≥Γ (G i−1)+1≥i.Now at
most 2 i−2(Γ(H)−1)+i−2 colours appear on G i−1[H]and at most 2 −2(Γ(H)−1)+1colours appear on
Si[H] by Proposition10 and because ci appears in all the H(v) for v ∈ Si. So in total at most 2 i−1 (Γ
(H) −1)+i−1 colours appear on Gi[H].
Corollary17
ψ (k,l) ≤2 k−1(l−1)+k−1.

(1)

If Γ (G) = 2 then Γ (G[H]) = 2k. (2)
23


ψ (2, k)=2k.

(3)

(ψ (3,2)=7.

(4)

Proof. (1) follows directly Theorem 16; Proposition 12 and Theorem 16 imply (2) and (3);
Proposition 15 and Theorem 16 yield (4).
Lemma18 Let α be a positive integer. If ψ (k,l) ≥ kl+ α then ψ(k’,l)≥ k’l+ α for all k’>k.
Proof. To prove this result it suffices to prove that if ψ(k,l)=kl+ α then ψ(k+1,l) ≥(k+1) l+α.
Then an easy induction will give the result.
Let G be a graph such that Γ (G [Kl]) = kl+ α. Let x be a vertex of G such that there exists a
greedy (kl +α)-colouring c such that the colour kl+ α appears on Kl(x). Let G1 and G2 be two
disjoint copies of G. For i= 1,2, we denote by vi the vertex vi ∈V (Gi) corresponding to v ∈V
(G). Let G’ be the graph obtained from G1+G2 by adding an edge between the two vertices x1
and x2.By Lemma 2 (1) and Proposition 6, Γ(G’) ≤Γ(G1+G2)+1=Γ(G)+1=k+1. Now let c’ be the
colouring of G’ [Kl] defined as follows:
- c’ (v1) =c (v) for v1∈V (G1[Kl]);
- c’ (v2) =c (v) for v2∈V (G2[Kl]){x2};
- The vertices of Kl(x2) are assigned distinct colours in {kl+α+1,...,(k+1)l+ α}.
Corollary 19 Let k ≥3 be an integer. Then ψ(k,p)≥(2k+1)p
Theorem 20 Let c≥1 be an integer. The following problem is CoNP-complete:
Instance: a graph G
Question: Γ (G) ≤cχ(G)?
Proof. Let G be a graph. If c1 is a colouring of G with t colours and c2 a greedy colouring of G
with more than ct colours, then the pair (c1,c2) forms a certificate that Γ(G)>cχ(G). Clearly, it can
be checked in polynomial time if a pair (c1,c2) is a certificate. So the problem is in CoNP.
Let us now show that this problem is CoNP-complete via a reduction to the problem of deciding
if Γ(G) ≤χ(G) for a given graph G, which is known to be CoNP-complete [15]. Let G be a graph.
Consider H = T2c [G]. Then χ(H)= 2χ(G) as ω(T2c)=χ(T2c)= 2. Moreover Γ (H) = 2c Γ (G) by
Theorem 13 (or Theorem 14). Hence Γ(H) ≤cχ(H) if and only if Γ(G)≤χ(G).
A similar proof yields that it is NP-complete to decide if Γ(G)≤ cω(G) as ω(T2c[G])= 2ω(G) and it
is CoNP-complete to decide if Γ(G)≤ω(G).
Theorem 21 Let c ≥1 be an integer. The following problem is CoNP-complete:
• Instance : a graph G.
• Question : Γ(G) ≤c ω(G)?

4. CARTESIAN PRODUCT
It is well-known that the chromatic number of the cartesian product of two graphs is the
maximum of the chromatic numbers of these graphs: χ(G□H)= max {χ(G),χ(H)}. Unfortunately,
no such formula holds for the grundy number. In this section, we are looking for bounds on the
grundy number of the cartesian product of two graphs in terms of the grundy numbers of these
graphs. We first show that such an upper bound does not exist. However, we show that for any
graph G there is a function hG such that for every graph H,Γ(G□H)≤hG(Γ(H)). Regarding lower
bounds, we give upper an lower bounds for the function
24


ϕ□ (k,l)= min{Γ(G□H)|Γ(G)=k and Γ(H)=l}
Let G and H be two graphs. For any v ∈V (G), the graph Hv of G□H induced by the vertices of
{v}×V(H) is isomorphic to H. Analogoulsy, for any x ∈V(H), the subgraph Gx of G□H induced
by the vertices of V(G) ×{x}is isomorphic to G.

4.1. Upper bounds
We denote by K p,p the complete bipartite graph with p vertices in each part.
Proposition 22 Let p≥2 be an integer. Then Γ(K p+1,p+1□K p+1,p+1) ≥ Γ(K p,p□K p,p)+1.
So Γ (K p,p□K p,p) ≥ p+1.
Proof. Let (X∪{x}, Y ∪{y}) be the bipartition of K p+1,p+1 with x /∈X and y /∈Y.
Then K p+1,p+1−{x,y} is a K p,p, so K p+1,p+1−{x,y}□K p+1,p+1−{x,y} is an induced K
p+1,p+1□K p+1,p+1.

p,p□K p,p

in K

Now the set ({x}×Y{y}) ∪ ({y}×X {x}) ∪ (X{x}×{x}) ∪ (Y{y}×{y}) is
a(X∪Y)×(X∪Y)dominating stable set. So by Proposition 5, Γ (K p+1,p+1□K p+1,p+1)≥Γ(K p,p□ K
p,p)+1.
As Γ (K 2,2)=2, an easy induction yields Γ(K p,p□ K p,p) ≥ p+1, for p≥2.
Proposition 23 Let G be a graph then for any positive integer k, hG(k) ≤Δ(G)·2 k−1+k. In other
words, for any graph H,Γ(G□H)≤Δ(G)·2 Γ(H)−1+Γ(H).
Proof. Let c be a greedy p-colouring of G□H. Let (v,x1) be a vertex coloured p=c1. For ever y
vertex x of H, set C(x):={c(w,x)|w ∈ NG(v)}. By extension, for every S∈V (H), we set C(S) =S x
∈S C(x). Let T1={x1}.
We have Γ (H<T1>)=1. Now, iteratively, as long as {1,...,p}C(Ti) ∪{c1,...,ci} is not empty, let us
construct T i+1 as follows. Let c i+1 be the largest integer of {1,...,p} C(Ti) ∪{c1,...,ci}.Then for
every x ∈Ti, the vertex (v,x) has a neighbour coloured c i+1 which by definition of C(x) is in Hv.
Hence there exists a stable set S i+1 of size at most |Ti| in H such that c(v,y)=c i+1 for every y ∈S i+1
and every vertex x ∈Ti has a neighbour in S i+1. Setting T i+1=Ti∪ S i+1, we have |T i+1|≤2|Ti|≤2i and
by Proposition 5, Γ (H<T i+1>) ≥i+1
Let i0 be the integer when the process terminates, i.e. when {1,..., p}=C(Ti0) ∪{c1,...,ci0}.
We have Γ (H) ≥Γ (H<Ti0>) ≥i0,|Ti0|≤2 i0−1 and |C(Ti0)|≤Δ(G)×|Ti0|. So p≤Δ(G)·2 i0−1+i0≤ Δ(G)·2
Γ(H)−1+Γ(H).
We think that the upper bound Δ (G) ·2
slightly better upper bounds.

k−1+k

is far to be tight. For some graphs one can get

Let us show an example when k=2. For a vertex v of graph G, we denote by d1G (v) or simply d1
(v) the maximum degree of a neighbour of v ,i.e. d1(v)=max{d(u)|u ∪N(v)}.
According to the proof of Theorem 25, p≤ max {dG(v)+d1G(v)+2 | v ∈V(G)}. We now show a
slightly better upper bound.
Proposition26 Let G be a graph. Then hG(2)≤max{min{2d(v)+2,2d1(v)+3}|v ∈V(G)}.
Proof. Let H be a complete bipartite graph and c be a greedy colouring of G□H with p colours.
Let x= (v, v’) be a vertex coloured with p and let (X,Y) be the bipartition of Hv with x ∈X.
25


Since x has dG(v) neighbours not in Hv, it has p−1−dG(v) neighbours in Y with distinct colours in
{1,...,p−1}. Let q be the largest integer in {1,...,p−1}that is assigned to a vertex in Y and let y be a
vertex coloured q. Then x has p−2−dG (v) neighbours in Y with distinct colours in {1,...,q−1}.
Now since y has at most dG(v) neighbours not in Hv, it has q−1−dG(v)neighbours in X with
distinct colours in {1,...,q−1}. As Hv is complete bipartite, the colours that appear on X do not
appear on Y.
Thus p −2−dG (v) +q−1−dG (v) ≤q−1, so p≤2dG (v) +2.
We claim that there is a vertex y=(u, u’) with u ∈NG(v)such that is assigned a colour p’≥ p−2 and
is adjacent to a vertex in Hv coloured p or p −1. Indeed x has a neighbour that is coloured
p−1.
If this neighbour is not in Hv it is the desired y. If not this neighbour z is in Y. Now both x and z
have a neighbour coloured p−2. But these two neighbours are not both in Hv otherwise they would
be adjacent. Hence one of them is not in Hv and is the desired y.
Now applying the same reasoning as above and taking into account that y has a neighbor outside
Hu with a larger colour than it’s, we obtain that p−2≤2dG (u) +1. So p≤2d1 (v) +3.
If the graph G has two adjacent vertices of G maximum degree then Proposition 26 yields the
same upper bound 2Δ(G)+2 as Theorem25. But for graphs in which vertices of high degree form
a stable set, this bound is far better. Consider for example a star K 1,p. By Proposition 26, for any
p≥ 2, h K 1,p (2) ≤5. Moreover K 1,p contains K 1,2 as an induced subgraph, so K 1,p□K 3,3 contains K
1,2□K 3,3 as an induced subgraph.

With similar arguments, one can improve a little bit the upper bound for hG for some graphs.
However, the upper bound is still exponential ink while we think hG is linear.
Conjecture 27 For any graph G, there is a constant CG such that h G (k) ≤ CG×k for any k.
A very first step to wards this conjecture would be to prove it for K 2. Baloghetal.[1]showed that h
K2(G) ≥ 2k because Γ(K2□K k[S2])= 2Γ(Kk[S2])= 2k with S2 the edgeless graph on two vertices.
They also conjectured that h K2 (G) ≥2k.
Denoting by Sk be the edgeless graph on k vertices, we now generalise both their conjecture and
their tightness examples.
Conjecture 28 Let k and n be two positive integers. Then h Kn(k)= n×k.
More generally, we conjecture the following:
Conjecture 29 For any graphs G and H, Γ (G□H) ≤ (Δ (H) +1) Γ (G)
If true these two conjectures would be tight as shown by the following proposition.
Proposition 30 Let k and n be two positive integers. Then Γ(Kn □ Kk [Sn])= n×k.
26


Proof. Δ (Kn □ Kk [Sn])=n×k−1 so Γ (Kn □Kk [Sn])≤n×k.
We now prove by induction on k that Γ (Kn □Kk [Sn])≥n×k. The result holds trivially when k=1.
Suppose now that k>1. Let us denote the vertices of Kk [Sn] by vi j,1≤i≤k,1≤ j≤ n so that for any
i the set {vi j |1≤ j≤n} is stable and the vertices of K n by x1,...,xn. Let T1= {(xj, v1j) |1≤ j≤n}.Then
T1 is a V (G) T1-dominating stable set. Indeed let (xj, vi l) be a vertex in V (G)T1. Then it is
adjacent to (xj, v1j) if i ± 1 and to (xl,v1 l) if i ≠ 1. More generally, for 1 ≤ i ≤ n, the set T i = {(xj,v1
i+j−1)| 1≤ j≤ n} is a V(G)Ti-dominating stable set. Note that (T1,...,Tn) is a partition of {(xj,v1
i)|1≤j≤n,1≤i≤n}and that Kn□K k[Sn is isomorphic to Kn□K k−1[Sn]. Hence applying Proposition 5,to
all the Ti one after another, we obtain Γ(Kn□Kk[Sn]) ≥ n+Γ (Kn□K k−1[Sn]). Now the induction
hypothes is yields Γ (Kn □ Kk [Sn]) ≥n×k.
Theorem 31 For any graph G, hG(k)≥Γ(G)+2k−2.
Proof. Set p=Γ (G) and n= p+2k−2. We will prove that Γ (Kk [Sn] □G)≥Γ(G)+2k−2. Let us denote
the vertices of Kk [Sn] by vij, 1≤ i≤ k, 1≤ j ≤ n so that for any i the set {vi j | 1≤ j ≤ n}.According to
Lemma 8, there are two vertices x and y that receive colour Γ (G) by some greedy colouring.
Observe that in Kk[Sn] □G the Gvj1,1≤ j≤ n are disjoint copies of G. Hence, by Lemma, we obtain
a partial greedy n-colouring of Kk [Sn] □G. So Γ(Kk [Sn] □G)≥n=Γ(G)+2k−2.

4.2. Lower bounds
As G and H are induced subgraphs of G□H then Γ (G□H) ≥max {Γ (G), Γ (H)}.Lemma 32 Let G
and H be two graphs. If χ (H) ≤Δ (G) then Γ (G□H) ≥Γ (H) +1.
Proof. W. l.o.g. we may assume that G and H have no isolated vertices. Let v be a vertex of G of
degree Δ (G) and let u1,..., u Δ(G) its neighbours. Let S1,...,S χ (H) be the stable set of a colouring of
H with χ (H) colours. The set ∪χ (H) i=1{ui} ×Si is a V (Hv)-dominating stable set. So by
Proposition 5, Γ (G□H) ≥Γ (H) +1.
Corollary33 Let G and H be two connected graphs such that Γ (G) =Γ (H) =k. Then Γ (G□H) ≥
k+1 unless G=H=K1 or G=H=K2.
Proof. If χ (H) ≤Δ (G) or χ(G)≤Δ(H), we have the result by Lemma 32. So we may assume that
(H)= χ(G)=Δ(G)+1=Δ(H)+1. Hence by Brooks Theorem [3], G and H are complete graphs or odd
cycles. If G = H= Kk, the result follows from Proposition 35 below.
If G and H are odd cycles, then one of P3□K2 and C3□K2 is an induced subgraph of G□H. These
graphs have grundy number 4; greedy 4-colourings are given Figure 3. So Γ (G□H) ≥4.
Lemma 32 yields a direct easy proof of a result of Hoffman and Johnson [9] stating that the
kdimensionnal hypercube Qk has Grundy number k+1 for k ≥3 and Γ (Q1) =Γ (Q2) =2. Recall
thatQ1=K2 and for k ≥2 then Qk=Q k−1□K2.
Proposition34 (Hoffman and Johnson [9]) for k ≥3, Γ (Qk) =k+1.
Proof. As Δ (Qk) =k, we have Γ (Qk) ≤k+1. Let us now prove the by induction that Γ (Qk) ≥k+1.
If k=3, a greedy 4-colouring is given. If k>3, then χ(K2) ≤Δ(Qk−1)=k. Hence by
Lemma33, Γ (Qk) ≥Γ (Qk−1) +1≥k+1
Proposition 35 For any p ≥2 then Γ (Kp□ Kp) =2p−2.

27


Corollary 33 implies that ϕ□ (k,k) ≥ k+1. To have better lower bound in ϕ□ (k,k), one may study
the function g(k)= min {Γ(G□G) |Γ(G)= k}. Clearly, g(2)= 2 and by Corollary 33 and Proposition
35, if k≥3, we have:
k ≤ϕ□ (k,k)≤g(k)≤2k−2.
Moreover every graph with grundy number 3 has either a K3 or a P4 as induced subgraph. But
Γ(K3□K3)= 4 by Proposition 35 and Γ(P4□P4)= 5 (As Δ(P4□P4)≤ 4 then Γ (P4□P4)≤ 5 and it is easy
to find a greedy 5-colouring of Γ(P4□P4). Hence g (3) =4.

5. DIRECT PRODUCT
A well known conjecture on graph colouring regards the chromatic number of the direct product
of graphs.
Conjecture36 (Hedetniemi[7])
χ (G×H)=min {χ(G), χ(H)}.
Poljak [10] proved that the function f defined by f(n)=min {χ(G×H) |χ(G)=χ(H)=n} is either
bounded by 9 or tends to infinity when n tends to infinity.
In this section, our aim is to find upper bounds of the grundy number of the direct product of two
graphs in terms of the grundy number of these graphs. Ideally, we would like to determine the
functions
ϕ× (k,l)=min {Γ(G×H)|Γ(G)=k and Γ(H)=l}.
Φ× (k,l)=max{Γ(G×H)|Γ(G)=k and Γ(H)=l}.
Let us first observe that if G=G1+G2 then G ×H=(G1×H)+(G2×H). Hence it is sufficient to
consider connected graphs. Furthermore, the direct product of a graph with K 1 is a graph without
any edge of order |G|. So ϕ×(k,1)=Φ×(k,1)=1. In the remaining of this section, all the graphs are
assumed to be connected of order at least 2. In particular, their grundy number is at least two.

5.1. Lower bounds
As every graph with grundy number k contains a k-atom as an induced subgraph then ϕ× (k,l)=
min {Γ(G×H) |G is a k-atom and H is an l-atom}. Furthermore if k≥k’ and l ≥l’ then ϕ× (k, l) ≥
ϕ× (k’, l’).
Theorem 37 Let G and H be two graphs with at least one edge. Then Γ (G×H) ≥Γ (G) +Γ (H) −2.
Hence if k ≥2and l≥2 then ϕ×(k,l)≥k+l−2.
28


Proof. Let k=Γ (G) and l=Γ (H). We prove the result by induction on k+l, the result holding
trivially if k=l=2.
Suppose now that k+l >4. Without loss of generality, we may assume that k ≥l. Let S 1,...,Sk be the
stable sets of a greedy p-colouring of G. Set G’= G−S1. Then S1 is a (V (G’))-dominating stable
set and Γ (G’) =k−1. Now, in G×H), the set S1×V (H) is V (G’×H)-dominating. Hence, by
Proposition5, Γ (G×H) ≥Γ (G ×H). Now, since Γ (G’) +Γ (H) =k+l−1, by induction hypothesis,
Γ (G’×H) ≥k+l−3. So Γ (G×H) ≥k+l−2.
This lower bound for ϕ k,l is attained when l=2 or k=l=3.
Corollary 38 For any integer k ≥2
ϕ× (k, 2)=k. (1)
ϕ×(3,3)=4. (2)
Proof. (1) The maximum degree of Kk ×K2 is k−1, so Γ (Kk×K2) ≤ k. So ϕ× (k, 2) ≤ k. But
Theorem37 yields ϕ× (k, 2) ≥k.
(2) One can easily check that Γ (P4×P4) = Γ (P4×C3) = Γ (C3×C3) = 4.

5.2. Upper bounds
Lemma 40 Let G and H be two graphs, and u and v be two vertices of G. If NG (u) = NG (v) then
Γ (G×H) = Γ ((G−u) ×H).
Proposition3 and Lemma40 directly imply:
Corollary 41 Let H be a graph such that Γ (H) =2. Then for every G, Γ (G×H) =Γ (G×K2).
In particular, Φ (k, 2) =max {Γ (G×K2)|Γ(G)=k} If G is bipartite then G×K2=G+G. So, by
Proposition6, Γ (G×K2) =Γ (G). Then Proposition3 yields Φ× (2, 2) = 2. There are non-bipartite
graphs G for which Γ (G×K2) =Γ (G). For example, K3×K2 is the 6-cycle so Γ (K3×K2) =3=Γ (K3).
There are also graphs G for which Γ (G×K2) = Γ (G): the jellyfish for example.
Definition 42 The head of the jellyfish J is the vertex h. Let G be a graph. Then the jellyfished of
G is the graph J(G) obtained from G by creating for each vertex v ∈V(G) a jellyfish J(v) whose
head is identified with v.
Proposition 43 Let G be a graph. Then Γ (J (G)) =Γ (G) +2.
Proof. Let c be a greedy colouring of J(G) with Γ(J(G)) colours. Let u be a vertex such that c (u)
≥4.
We claim that u is in V (G). Suppose not. Let v be the vertex of G such that u ∈J(v). Then u must
be the vertex of J (v) of degree 3 adjacent to v. Since u has a neighbour of each colour smaller
than 4 the vertex v must be assigned 3.
But then the two others neighbours of u are coloured 1, a contradiction.
Now as Γ (J−h) =2 the neighbours of u coloured 3 are in G. Therefore the colouring c’ defined on
S= {v∈V(G) | c(v)≥ 3} by c’(v)= c(v)−2 is a greedy colouring of G<S> with Γ(J(G))−2 colours.
So Γ (J (G)) ≤Γ (G) +2.
Lemma44 If G ×K2 contains an induced binomial tree of index k then there is a graph H such that
H ×K2 contains an induced binomial tree of index k+3and Γ(H)=Γ(G)+2.

29


Proof. Let T be an induced Tk of G ×K2. Let S={v ∈ V(G) | {(v,1),(v,2)}⊂V(Tk)}. Let G’ be the
graph obtained from G by blowing up each vertex v of S with a stable set of size two {v1,v2}.
By Proposition 2 (3), Γ (G’) =Γ (G).
Set H= J (G’). By Proposition 43, Γ (H) =Γ (G’) +2= Γ (G) +2.
Let us now show that H ×K2 contains a T k+3. By construction, the subgraph of G’×K2 induced by
(V (Tk) (S×K2)) ∪v∪ S {(v1, 1), (v2, 2)} is aTk. Note that for every vertex v ∈ V (G’) at most one
of {(v, 1), (v, 2)} is in V(T’). Now every vertex of T’ is the root of a T 3 in its associated J×K2. All
theseT3 together with T’ form an induced T k+3 of H ×K2.
Corollary45 Φ× (2k+1, 2) ≥3k+1and Φ× (2k, 2) ≥3k−1.
Note that Corollary 41 may not be generalised to graphs H such that Γ (H) = 3. Indeed 4=
Γ(G3×K2)=Γ(K3×K2)=3

6. CONCLUSIONS
The Grundy number of a graph G, denoted by Γ(G), is the largest k such that G has a greedy
kcoloring, that is a coloring with colours obtained by applying the greedy algorithm according to
some ordering of the vertices of G. We study the Grundy number of the lexicographic, Cartesian
and direct products of two graphs in terms of the Grundy numbers of these graphs .We give the
lower and the upper bounds of the Grundy number for many graphs.

ACKNOWLEDGMENTS
I thank Mr. Fréderic Havet for his collaboration.

REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]

[15]

J.Balogh,S.G.Hartke and Q.LiuandG.Yu. On the first-fit chromatic number of graphs. SIAMJ.
Discrete Math.22(3):887–900, 2008.
J. A. Bondy and U. S. R. Murty.Graphtheory, GTM 244, 2008
R. L. Brooks. On colouring the nodes of a network. Proc. Cambridge Phil. Soc. 37:194– 97,1941
T.Emden-Weinert, S.Hougardy, and B. Kreuter. Uniquely colourable graphs and the hardness of
colouring graphs of large girth. Combin. Probab. Comput.7(4):375–386, 1998.
D. Geller and S. Stahl. The chromatic number and other parameters of the lexicographic product.
J. Combin. Theory Ser. B19:87–95, 1975
A.Gy´arf´asandJ.Lehel. n-line and first-fit coloring of graphs. J.GraphTheory12:217–227, 1988
S. Hedetniemi. Homomorphism of graphs and automata. Univ. Michigan Technical Report, 1966
S.M. Hedetniemi, S.T. Hedetniemi and A. Beyer. A linear algorithm for the Grundy (coloring)
number of a tree. Congr. Numer. 36:351–363, 1982.
D.G.Hoffman and P.D.JohnsonJr. Greedy colorings and the Grundy chromatic number of the ncube.
Bull. Inst. Combin. Appl.26:49–57, 1999
S. Poljak. Coloring digraphs by iterated antichains. Comment. Math. Univ. Carolin. 32:209212,
1991.
S. Stahl. n-tuple colorings and associated graphs.J. Combin. Theory Ser. B20:185–203, 1976.
M. Aste , F. Havet and C. l.Sales ThemeCOM — Systemes communicants Equipe-Projet
Mascotte , Octobre 2008.
S. Klavzar. Coloring graph products. Discrete Math.155:135–145, 1996.
M.MolloyandB.Reed. Colouring graphs when the number of colours is nearly the maximum
degree. In Proceedings of the Thirty-Third Annual ACM Symposium on Theory of Computing
pages 462–470 (electronic), New York, 2001. ACM
M. Zaker. Results on the Grundy chromatic number of graphs. Discrete Math. 306(23):3166–
)=^3173, 2006.
30

On the equality of the grundy numbers of a graph

More Related Content

What's hot (19)

Viewers also liked (13)

Similar to On the equality of the grundy numbers of a graph (20)

More from ijngnjournal (20)

Recently uploaded (20)

On the equality of the grundy numbers of a graph