Radio k-Labelings for Cartesian Products of Graphs

Mustapha Kchikech, Riadh Khennoufa and Olivier Togni

LE2I, UMR CNRS 5158

Université de Bourgogne, 21078 Dijon cedex, France
{kchikech, khennoufa, otogni}@u-bourgogne.fr

Abstract
Frequency planning consists in allocating frequencies to the trans-
mitters of a cellular network so as to ensure that no pair of transmitters
interfere. We study the problem of reducing interference by modeling
this by a radio k-labeling problem on graphs: For a graph G and an in-
teger k ≥ 1, a radio k-labeling of G is an assignment f of non negative
integers to the vertices of G such that

|f (x) − f (y)| ≥ k + 1 − dG (x, y),

for any two vertices x and y, where dG (x, y) is the distance between
x and y in G. The radio k-chromatic number is the minimum of
max{f (x) − f (y) : x, y ∈ V (G)} over all radio k-labelings f of G.
In this paper we present the radio k-labeling for the Cartesian prod-
uct of two graphs, providing upper bounds on the radio k-chromatic
number for this product. These results help to determine upper and
lower bounds for radio k-chromatic numbers of hypercubes and grids.
In particular, we show that the ratio of upper and lower bounds of
the radio number and the radio antipodal number of the square grid
is asymptotically 23 .

Keywords: Graph theory, radio channel assignment, radio k-labeling,

Cartesian product, radio number, antipodal number.

1 Introduction
In wireless networks, an important task is the management of the radio
spectrum, that is the assignment of radio frequencies to transmitters in a
way that avoid interferences. Interferences can occur if transmitters with
close locations receive close frequencies. The problem, often modeled as
a coloring problem on the graph where vertices represent by transmitters
and edges indicate closeness of the transmitters, has been studied by several
authors under different scenarios.

1
 In this paper, we study the radio k-labeling problem defined by Char-
trand et al. [2, 3]. Formally, for a graph G = (V, E), we denote by dG (x, y)
the distance between two vertices x and y, and by D(G) the diameter of G.
A radio k-labeling of G is a function f : V → N such that for every two
distinct vertices x and y of G the following is satisfied :

|f (x) − f (y)| ≥ k + 1 − dG (x, y).

The span of the function f denoted by λk (f ), is max{f (x) − f (y) : x, y ∈

V (G)} . The radio k-chromatic number λk (G) of G is the minimum span of
all radio k-labelings of G.
Determining the radio k-chromatic number seems to be a difficult task,
even for particular graphs. For instance, the radio k-chromatic number
for paths was studied in [3, 6], where lower and upper bounds were given.
Radio labelings for particular values of k were also considered: the radio
k-chromatic number of paths and cycles is studied for k = D(G) − 1 in
[1, 2] known as radio antipodal number and for k = D(G) in [12] known
as radio number. The radio number of trees and square of paths has been
studied in [10, 11]. Recently, the radio antipodal number of the hypercube
was determined [8].
Notice that, although the authors in [3] only consider radio k-labelings
for k ≤ D(G), one can also consider the case k > D(G). The motivation
behind the study of the case k > D(G) is of two kinds: first this case seems
less difficult to study than the case k ≤ D(G) and secondly, computing the
radio k-chromatic number of a graph for k ≥ D(G) can help to compute the
radio k-labeling number of other graphs with larger diameter, as it is done
in [6].
For the Cartesian product, we use the notation from [5] : The Cartesian
product G2G0 of two graphs G and G0 is the graph with vertex set V (G) ×
V (G0 ) and edge set {(xi , uj )(xi0 , uj 0 )| i = i0 and uj uj 0 ∈ E(G0 ), or xi xi0 ∈
E(G) and j = j 0 }. Therefore, to each vertex uj of G0 corresponds a copy
Gj of G in G2G0 , with 1 ≤ j ≤ |V (G)|.
As shown in [6], the radio k-chromatic number is related to two other
graph parameters: the upper Hamiltonian number [9] of a graph G, denoted
n−1
by h+ (G), is the maximum of
P
dG (π(i + 1), π(i)), over all cyclic permuta-
i=0
tions π of the vertices of G. The upper traceable number, denoted by t+ (G),
is obtained from h+ (G) by ignoring the distance between the first and the
n−2
last vertex: t+ (G) = max
P
dG (π(i + 1), π(i)).
π i=0
Král et al. in [9] showed that the problem of determining the upper
Hamiltonian number of a graph is N P -hard. The same method can be used
to prove that computing the upper traceable number is also an N P -hard
problem.

2
 We shall use the following results from [6], which present a lower bound
and an upper bound on the radio k-chromatic number of a graph G in terms
of the parameter t+ (G) and of the chromatic number χ(Gk ) of the k th power
Gk of G (i.e. the graph with the same vertex set as G and with edges between
vertices at distance at most k in G).
Theorem 1 ([6]). Let G be a graph of order n, then for any positive integer
k,
λk (G) ≥ (n − 1)(k + 1) − t+ (G);
Moreover, if k ≥ 2D(G) − 2, then

λk (G) = (n − 1)(k + 1) − t+ (G).

Theorem 2 ([6]). For any graph G and any integer k ≥ 1,

λk (G) ≤ k(χ(Gk ) − 1).

The aim of this paper is to find relations between the radio k-chromatic
number of the Cartesian products of two graphs, and some (other) coloring
parameters on the factors. In Section 2, we propose general upper bounds for
the radio k-chromatic number for the product of two graphs. In Section 3,
we find more refined results for the product of a graph and a path. Applying
these results, we present in Section 4 upper and lower bounds for the radio
k-chromatic number of the hypercube and of the grid. In particular, for the
radio antipodal and radio numbers of the grid, the ratio of the upper and
lower bounds is small (asymptotically equal to 23 ).

2 Radio k-labelings of the Cartesian product of

two graphs
In this section we give general bounds for the radio k-chromatic number of
the Cartesian product of two graphs G and G0 .
Theorem 3. For any two graphs G and G0 of order n ≥ 2 and m respec-
tively, and for any integer k ≥ D(G2G0 ) − 1,

λk (G2G0 ) ≤ mλk (G) + (m − 1)k − t+ (G0 ).

Proof. Let f be a radio k-labeling of G with λk (f ) = λk (G).

Let x0 , x1 , . . . , xn−1 be an ordering of the vertices of G such that f (xi ) ≤
f (xi+1 ) and let u0 , u1 , . . . , um−1 be an ordering of the vertices of G0 such
m−2
dG0 (ui+1 , ui ) = t+ (G0 ).
P
that
i=0
For each vertex ui of G0 we associate a copy Gi of G in G2G0 , where
V (Gi ) = {Xi0 , Xi1 , . . . , Xin−1 } with Xij = (xj , ui ). With this notation, we
0
have dG2G0 (Xij , Xij0 ) = dG (xj , xj 0 ) + dG0 (ui , ui0 ).

3
 Therefore, we can define a labeling g of G2G0 by setting

 j
 g(X0 ) = f (xj ), for 0 ≤ j ≤ n − 1,
0 n−1 n−1
g(Xi ) = g(Xi−1 ) + k + 1 − dG2G0 (Xi0 , Xi−1 ), for 1 ≤ i ≤ m − 1,
g(Xij ) = g(Xi0 ) + g(X0j ), for 1 ≤ j ≤ n − 1; 1 ≤ i ≤ m − 1.


The maximum label used is

n−1
g(Xm−1 0
) = g(Xm−1 )+g(X0n−1 ) = g(Xm−2
n−1 0
)+λk (G)+k+1−dG2G0 (Xm−1 n−1
, Xm−2 ),

and thus
m−2
X
n−1
g(Xm−1 ) = mλk (G) + (m − 1)(k + 1) − 0
dG2G0 (Xp+1 , Xpn−1 ).
p=0

m−2 m−2
0 , X n−1 ) ≥
P P
As n ≥ 2, then dG2G0 (Xp+1 p (1 + dG0 (up+1 , up )) = m − 1 +
p=0 p=0
t+ (G0 ) because x0 6= xn−1 . Consequently
n−1
g(Xm−1 ) ≤ mλk (G) + (m − 1)k − t+ (G0 ).

Now, we show that g is a radio k-labeling of G2G0 by checking the distance

condition for each pair of vertices in G2G0 : we want
0 0
|g(Xij ) − g(Xij0 )| ≥ k + 1 − dG2G0 (Xij , Xij0 ).

Case 1: If the two vertices are in the same copy Gi of G2G0 , then the
difference between their labels given by g is the same as that between the
corresponding two vertices in G0 :
0 0
|g(Xij ) − g(Xij )| = |g(Xi0 ) + g(X0j ) − (g(Xi0 ) + g(X0j ))|
0
= |g(X0j ) − g(X0j )|
= |f (xj ) − f (xj 0 )|
≥ k + 1 − dG (xj , xj 0 ).
0
As dG2G0 (Xij , Xij ) = dG (xj , xj 0 ), we obtain
0 0
|g(Xij ) − g(Xij )| ≥ k + 1 − dG2G0 (Xij , Xij ).

Case 2: If the two vertices are not in the same copy of G2G0 , then we just
j 0
check the distance condition between two vertices Xi+1 and Xij which are

4
in two successive copies Gi+1 and Gi respectively. We have
j 0 0
|g(Xi+1 0 ) + g(X j ) − g(X j )|
) − g(Xij )| = |g(Xi+1 0 i
0
= |g(Xin−1 ) + k + 1 − dG2G0 (Xi+1 0 , X n−1 ) + g(X j ) − g(X j )|
i 0 i
0
≥ g(Xin−1 ) − g(Xij ) + g(X0j ) (because we have k ≥ D(G2G0 ) − 1)
0
≥ g(X0n−1 ) − g(X0j ) + g(X0j )
0
≥ g(X0j ) − g(X0j )
≥ f (xj ) − f (xj 0 )
≥ k + 1 − dG (xj , xj 0 )
j 0
≥ k + 1 − dG2G0 (Xi+1 , Xij ).

Thus g is a radio k-labeling and λk (G2G0 ) ≤ mλk (G)+(m−1)k−t+ (G0 ).

We now give another upper bound for the radio k-chromatic number of
the Cartesian product which is valid for any k ≥ 2.
Theorem 4. For any two graphs G and G0 of order n and m respectively
and for any integer k ≥ 2,

λk (G2G0 ) ≤ χ(G0k )(λk (G) + k − 1) − k + 1.

Proof. Let f be a radio k-labeling of G with λk (f ) = λk (G) and let x0 , x1 , . . . , xn−1

be an ordering of the vertices of G such that f (xi ) ≤ f (xi+1 ).
Let c be a proper vertex-coloring of G0k with colors from {0, 1, . . . , χ(G0k )−
1}. Denote the vertices of G0 by uj , 0 ≤ j ≤ m − 1.
Consider the labeling g of G2G0 given by

g((xi , uj )) = f (xi ) + c(uj )(λk (G) + k − 1).

The maximal label used by g is χ(G0k )(λk (G) + k − 1) − k + 1. To show

that g is a radio k-labeling of G2G0 , we have to check that the distance
condition is satisfied.
Notice that for any two vertices in the same copy Gj of G, the condition
can be easily verified.
Next, two vertices (xi , uj ) and (xi0 , uj 0 ) with c(uj ) = c(uj 0 ) are at dis-
tance at least k + 1 in G2G0 (since c is a proper coloring of G0k ), thus the
condition is also verified for these two vertices.
Finally, for two vertices (xi , uj ) and (xi0 , uj 0 ) with c(uj ) 6= c(uj 0 ) and
i > i0 , we have

|g((xi , uj )) − g((xi0 , uj 0 ))| = |f (xi ) − f (xi0 ) + (c(uj ) − c(uj 0 ))(λk (G) + k − 1)|.

If c(uj ) > c(uj 0 ) then

|f (xi ) − f (xi0 ) + (c(uj ) − c(uj 0 ))(λk (G) + k − 1)| ≥ f (xi ) − f (xi0 )

≥ k + 1 − dG (xi , xi0 )
≥ k + 1 − dG2G0 ((xi , uj ), (xi0 , uj 0 )).

5
 If c(uj ) < c(uj 0 ) then
|f (xi ) − f (xi0 ) + (c(uj ) − c(uj 0 ))(λk (G) + k − 1)| = (c(uj 0 ) − c(uj ))(λk (G) +
k − 1) − f (xi ) + f (xi0 ), because λk (G) + k − 1 − f (xi ) + f (xi0 ) ≥ 0.
Thus,
|g((xi , uj )) − g((xi0 , uj 0 ))| ≥ λk (G) + k − 1 − (f (xi ) − f (xi0 ))
≥ k + 1 − dG2G0 ((xi , uj ), (xi0 , uj 0 )) + dG2G0 ((xi , uj ), (xi0 , uj 0 )) − 2
+λk (G) − (f (xi ) − f (xi0 ))
≥ k + 1 − dG2G0 ((xi , uj ), (xi0 , uj 0 ))
since dG2G0 ((xi , uj ), (xi0 , uj 0 )) ≥ 2 and f (xi ) ≤ λk (G).

3 Radio k-labelings of the products of a graph and

a path
In [12], the value of t+ (G) was determined for the path.
Lemma 1 ([12]). For any integer n ≥ 2,
 1 2
+
t (Pn ) = 2n − 1 if n is even,
1 2 − 1) − 1 if n is odd.
2 (n
Then, applying Theorem 3, we obtain:
Theorem 5. For any graph G of order n and for any integers m and k,
with k ≥ D(G) + m − 2,
mλk (G) + (m − 1)k − 21 (m2 − 1) + 1 if m is odd,

k
λ (G2Pm ) ≤
mλk (G) + (m − 1)k − 12 m2 + 1 if m is even.
Another bound for the product of a graph and a path can be determined
directly:
Theorem 6. For any graph G and for any integer k ≥ 2,
λk (G2Pm ) ≤ k(λk (G) + min{k, m − 1}).
Proof. Let f be a radio k-labeling of G such that λk (G) = λk (f ). Let
(x0 , x1 , ..., xn−1 ) be an ordering of the vertices of G such that f (xi ) ≤ f (xi+1 )
and let V (Pm ) = {0, 1, . . . , m − 1}.
Now, we construct a labeling g of G2Pm :
g((xi , j)) = k(f (xi ) + (j mod k))
In order to show that g is a radio k-labeling, first note that for any j, j 0 and
for any xi , xi0 , we have
dG2Pm ((xi , j), (xi0 , j 0 )) = dG (xi , xi0 ) + |j − j 0 |.
Hence, if |j − j 0 | ≥ k, then the condition |g((xi , j)) − g((xi , j 0 ))| ≥ k +
1 − dG2Pm ((xi , j), (xi , j 0 )) is satisfied provided that (xi , j) 6= (xi , j 0 ).

6
 • For any j 6= j 0 , |j − j 0 | < k and for any xi , we have

|g((xi , j))−g((xi , j 0 ))| = k|(j mod k)−(j 0 mod k)| ≥ k ≥ k+1−dG2Pm ((xi , j), (xi , j 0 )).

• For any j and for any i > i0 , we have

|g((xi , j)) − g((xi0 , j))| = k|f (xi ) − f (xi0 )|
≥ |f (xi ) − f (xi0 )|
≥ k + 1 − dG (xi , xi0 )
≥ k + 1 − dG2Pm ((xi , j), (xi0 , j)).

• For any j 6= j 0 , |j − j 0 | < k, let jk = (j mod k) and jk0 = (j 0 mod k).

For any i > i0 , we have
|g((xi , j)) − g((xi0 , j 0 ))| = k|f (xi ) − f (xi0 ) + jk − jk0 |.
If jk > jk0 then,
|g((xi , j)) − g((xi0 , j 0 ))| = k(f (xi ) − f (xi0 ) + jk − jk0 )
≥ f (xi ) − f (xi0 )
≥ k + 1 − dG (xi , xi0 )
≥ k + 1 − dG2Pm ((xi , j), (xi0 , j 0 )).
If jk < jk0 then,
if f (xi ) − f (xi0 ) ≥ jk0 − jk , we have,
|g((xi , j)) − g((xi0 , j 0 ))| = k(f (xi ) − f (xi0 ) − (jk0 − jk ))
≥ f (xi ) − f (xi0 ) − (jk0 − jk )
≥ k + 1 − dG (xi , xi0 ) − (jk0 − jk )
≥ k + 1 − dG (xi , xi0 ) − (j 0 − j)
≥ k + 1 − dG2Pm ((xi , j), (xi0 , j 0 )).
If f (xi ) − f (xi0 ) ≤ jk0 − jk , then
|g((xi , j)) − g((xi0 , j 0 ))| = k(jk0 − jk − (f (xi ) − f (xi0 )))
≥ jk0 − jk − (f (xi ) − f (xi0 ))
≥ f (xi ) − f (xi0 ) − (jk0 − jk )
≥ k + 1 − dG (xi , xi0 ) − (jk0 − jk )
≥ k + 1 − dG (xi , xi0 ) − (j 0 − j)
≥ k + 1 − dG2Pm ((xi , j), (xi0 , j 0 )).

Thus, for any j, j 0 and for any xi , xi0 , we have

|g((xi , j)) − g((xi0 , j 0 ))| ≥ k + 1 − dG2Pm ((xi , j), (xi0 , j 0 )).
Consequently, g is a radio k-labeling of G2Pm and
λk (G2Pm ) ≤ λk (g) ≤ k(λk (G) + m − 1).

7
4 Radio k-labelings for hypercubes and grids
4.1 Hypercubes
Let Hn be the hypercube of dimension n (Hn = P2 2 · · · 2P2 ).
| {z }
n times

Lemma 2. For any integer n ≥ 2,

t+ (Hn ) = 2n−1 (2n − 1) − (n − 1).

Proof. First, observe that there are no three vertices x, y, z in Hn such that
dHn (x, y) = dHn (y, z) = n. Thus, the best we can do in order to maximize
the sum of distances between consecutive vertices is to find an ordering of the
vertices such that the distance between consecutive vertices is alternately n
and n − 1.
This can be easily done by considering an Hamiltonian path P of a
subgraph H of Hn isomorphic to Hn−1 (it is straightforward that such a
path exists). The ordering is obtained by taking the first vertex of P , then
its antipodal vertex, then the second vertex of P , its antipodal, and so on.
Therefore,
n−1 2n−1
2X X−1
+
t (Hn ) = n+ (n − 1) = 2n−1 (2n − 1) − (n − 1).
i=1 i=1

Theorem 7. For the hypercube Hn of dimension n ≥ 2 and for any k ≥ 2,

(2n − 1)k − 2n−1 (2n − 3) + n − 2 ≤ λk (Hn ) ≤ (2n − 1)k − 2n−1 + 1.

Moreover, for k ≥ 2n − 2,

λk (Hn ) = (2n − 1)k − 2n−1 (2n − 3) + n − 2.

Proof. The lower bound is a direct consequence of Theorem 1 and Lemma 2.

To show the upper bound, observe that Hn = Hn−1 2P2 . Thus, applying
Theorem 3 inductively, we obtain

λk (Hn ) = λk (Hn−1 2P2 )

≤ 2λk (Hn−1 ) + k − 1
≤ 22 λk (Hn−2 ) + (k − 1)(2 + 1)
...
n−2
≤ 2n−1 λk (P2 ) + (k − 1)
P i
2
i=0
≤ (2n − 1)k − 2n−1 + 1.

8
4.2 Grids
Let Mm,n = Pm 2Pn denote the 2-dimensional grid.
In this section, we provide upper and lower bounds for the radio k-
chromatic number for the grid Mm,n only in terms of k as given in [3, 6] for
the path.
The k-distance chromatic number of a 2-dimensional grid (or equiva-
lently, the chromatic number of the k th power of the 2-dimensional grid)
was determined in [4]. Using Theorem 2, we obtain an upper bound as
given in the result below:
Theorem 8. For the grid Mm,n and for any integer k > 2,
 1 3 2
k
λ (Mm,n ) ≤ 2 (k + 2k − k) if k is odd,
1 3 2)
2 (k + 2k if k is even.

In order to present a lower bound, we need the following lemma.

Lemma 3. For any integer p ≥ 1,

h+ (M2p,2p ) ≤ 8p3 .

Proof. Let V (M2p,2p ) = {(i, j)|0 ≤ i, j ≤ 2p − 1}.

Let N = 4p2 and let X0 , · · · , XN −1 be an ordering of V (M2p,2p ) such that
−1
NP
h+ (M2p,2p ) = dM2p,2p (XJ , XJ+1 ). Remark that XJ+1 = (i, j) and XJ =
J=0
(i0 , j 0 ) for some 0 ≤ i, i0 , j, j 0 ≤ 2p − 1. Thus, dM2p,2p (XJ , XJ+1 ) = |i − i0 | +
NP−1
|j − j 0 |. Therefore, the summation dM2p,2p (XJ , XJ+1 ) consists in 4p2
J=0
distances, each distance corresponding to 4 numbers (two with a positive sign
and two with a negative sign). Moreover, each number i ∈ {0, 1, · · · , 2p − 1}
occurs 8p times in the summation.
Consequently, the configuration achieving the maximum summation is
when each number i ∈ {0, 1, . . . , p − 1} occurs 8p times as −i and each of
{p, p + 1, . . . , 2p − 1} occurs 8p times as i. In that case we obtain
−1
NP 2p−1
P p−1
P
dM2p,2p (XJ , XJ+1 ) ≤ 8p( i− i)
J=0 i=p i=0
p−1
P
= 8p( p)
i=0
= 8p3

Theorem 9. For any positive integer k and for any integer p ≥ 1,

λk (M2p,2p ) ≥ 4p2 (k − 2p + 1) − k.

9
Proof. The result follows by combining Theorem 1 and Lemma 3, using the
fact that t+ (M2p,2p ) ≤ h+ (M2p,2p ) − 1.

Theorem 10. For the grid Mm,n and for any integer k ≥ 2 with min{m, n} ≥
2d k3 e,
4 16
λk (Mm,n ) ≥ (k 3 + 3k 2 ) − k − .
27 27
Proof. It is easily seen that for any integers m, n, m0 , n0 such that m0 ≥ m
and n0 ≥ n then λk (Mm0 ,n0 ) ≥ λk (Mm,n ) since any radio k-labeling of Mm0 ,n0
provides a radio k-labeling of Mm,n .
The lower bound is obtained by setting p = d k3 e in Theorem 9 (we chose
this value of p in order to maximize the expression 4p2 (k − 2p + 1) − k).
4 2
If k ≡ 0 mod 3 then p = k3 and λk (M2p,2p ) ≥ 27 k (k + 3) − k ≥ 274
(k 3 +
3k 2 ) − k − 16
27 .
If k ≡ 1 mod 3 then p = k+2 k 4 2
3 and λ (M2p,2p ) ≥ 27 (k + 2) (k − 1) − k =
4 3 2 16
27 (k + 3k ) − k − 27 .
If k ≡ 2 mod 3 then p = k+1 k 4 3 4
3 and λ (M2p,2p ) ≥ 27 (k + 1) − k ≥ 27 (k +
3
2 16
3k ) − k − 27 .

Using Theorem 5 and the following theorem in [6], we obtain another

upper bound for the grid Mn,n .

Theorem 11 ([6]). For the path Pn of order n ≥ 2,

(n − 1)k − 21 (n − 1)2 + 1 if n is odd and k ≥ n − 2,


k
λ (Pn ) =
(n − 1)k − 12 n(n − 2) if n is even and k ≥ n − 1.

Proposition 1. For the grid Mn,n = Pn 2Pn of order n2 and for any k ≥
2n − 3,

(n − 1)k − 21 (n − 1)(n2 − 1) + 2 if n is odd,

 2
λk (Mn,n ) ≤
(n2 − 1)k − 12 n2 (n − 1) + 1 if n is even.

In particular, for k = D(Mn,n ) − 1 and k = D(Mn,n ) the above results

give bounds for the radio antipodal number and radio number of Mn,n . The
ratio of the upper and lower bounds is asymptotically equal to 32 .

Proposition 2. The radio antipodal number of the mesh Mn,n satisfies

3 5 3 9
n3 − 3n2 + n + 2 ≤ λ2n−3 (Mn,n ) ≤ n3 − n2 − n + .
2 2 2 2
Proposition 3. The radio number of the mesh Mn,n satisfies

3 7
n3 − 2n2 − n + 2 ≤ λ2n−2 (Mn,n ) ≤ (n3 − n2 − n + ).
2 3

10
Conclusion
We have presented several bounds for the radio k-chromatic number of the
Cartesian product of graphs. Although it seems difficult to judge sharpness
of the bounds, we have shown in Section 4 that for some values of k near
the diameter of the graph, some of the bounds proposed are relatively close
to the optimal. Moreover, this study is among the first to consider radio k-
labeling of graphs different from a path or a cycle. An interesting question
is to know if the radio k-chromatic number of a graph G is closer to the
chromatic number of the graph Gk than to k times this number.

Acknowledgements
We wish to thank Daphne Der Fan Liu for helpful discussions and for pro-
viding us the notion of upper Hamiltonian number of a graph.

References
[1] G. Chartrand, D. Erwin, and P. Zhang. Radio antipodal colorings of
cycles. Congr. Numer., 144:129-141, 2000.
[2] G. Chartrand, D. Erwin, and P. Zhang. Radio antipodal colorings of
graphs. Math. Bohem., 127(1):57–69, 2002.
[3] G. Chartrand, L. Nebeský and P. Zhang. Radio k-colorings of paths.
Discussiones Mathematicae Graph Theory, 24:5-21, 2004.
[4] G. Fertin, E. Godard, A. Raspaud. Acyclic and k-distance coloring of
the grid. Inform. Process. Lett., 87(1):51–58,2003.
[5] W. Imrich and S. Klavžar. Product graphs. Structure and recognition,
With a foreword by Peter Winkler. Wiley-Interscience Series in Discrete
Mathematics and Optimization. Wiley-Interscience, New York, 2000.
[6] M. Kchikech, R. Khennoufa and O. Togni Linear and cyclic radio
k-labelings of trees. Discussiones Mathematicae Graph Theory, 27
(1):105-123, 2007.
[7] R. Khennoufa and O. Togni. A note on radio antipodal colourings of
paths. Math. Bohemica, 130 (3):277-282, 2005.
[8] R. Khennoufa and O. Togni. The Radio Antipodal Number of the
Hypercube. Submitted, 2007.
[9] D. Král, L.-D. Tong and X. Zhu. Upper Hamiltonian numbers and
Hamiltonian spectra of graphs. Australasian Journal of Combinatorics,
35:329-340, 2006.

11
[10] D. Liu. Radio Number for Trees. manuscript, 2006.

[11] D. Liu and M. Xie. Radio Number for Square Paths. Discrete Math.,
to appear.

[12] D. Liu and X. Zhu. Multi-level distance labelings for paths and cycles.
SIAM J. Discrete Mathematics, 19 : 610-621, 2005.

12