• 沒有找到結果。

Distance-two labelings of graphs

N/A
N/A
Protected

Academic year: 2021

Share "Distance-two labelings of graphs"

Copied!
6
0
0

加載中.... (立即查看全文)

全文

(1)

www.elsevier.com/locate/ejc

Distance-two labelings of graphs

Gerard J. Chang

a

, Changhong Lu

b,1 aDepartment of Mathematics, National Taiwan University, Taipei 106, Taiwan

bDepartment of Mathematics, Hunan Normal University, Changsha 410081, People’s Republic of China

Received 25 May 2001; accepted 26 September 2002

Abstract

For given positive integers j ≥ k, an L( j, k)-labeling of a graph G is a function f : V (G) → {0, 1, 2, . . .} such that | f (u) − f (v)| ≥ j when dG(u, v) = 1 and | f (u) − f (v)| ≥ k when dG(u, v) = 2. The L( j, k)-labeling number λj,k(G) of G is defined as the minimum m such that there is an L( j, k)-labeling f of G with f (V (G)) ⊆ {0, 1, 2, . . . , m}. For a graph G of maximum degree∆ ≥ 1 it is the case that λj,k(G) ≥ j + (∆ − 1)k. The purpose of this paper is to study the structures of graphs G with maximum degree∆ ≥ 1 and λj,k(G) = j + (∆ − 1)k.

c

2003 Elsevier Science Ltd. All rights reserved.

Keywords: Distance; Labeling; Degree; Tree; Star; Algorithm; Neighbor

1. Introduction

The problem of vertex labeling with a condition at distance two, proposed by Griggs and Roberts [16], arose from a variation of the channel assignment problem introduced by Hale[10]. Suppose a number of transmitters are given. We must assign a channel to each of the given transmitters such that the interference is avoided. In order to reduce the interference, any two “close” transmitters must receive channels at least k apart, and any two “very close” transmitters must receive channels at least j apart, where j ≥ k are two given positive integers. One can construct an interference graph for this problem so that the transmitters are the vertices and there is an edge joining two “very close” transmitters. Two transmitters are defined as “close” if the corresponding vertices are of distance two.

Then, for a given graph G, an L( j, k)-labeling is defined as a function f : V (G) → {0, 1, 2, . . .} such that | f (u) − f (v)| ≥ j when dG(u, v) = 1 and | f (u) − f (v)| ≥ k when dG(u, v) = 2, where dG(u, v), the distance of u and v, is the minimum length of a path between u andv. The L( j, k)-labeling number λj,k(G) of G is the smallest number

1 Permanent address: Department of Mathematics, East China Normal University, Shanghai 200062, P. R. China.

0014-5793/03/$ - see front matter c 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0195-6698(02)00134-8

(2)

λj, j(T ) = j∆. For j = 2 and k = 1, Griggs and Yeh

means of a first-fit (greedy) algorithm: for any tree T with maximum vertex degree∆ ≥ 1,

∆ + 1 ≤ λ2,1(T ) ≤ ∆ + 2.

This result gives rise to the question of classifying those trees withλ2,1(T ) = ∆ + 1 or alternatively∆ + 2. Although Griggs and Yeh conjectured that the classification problem is NP-complete, Chang and Kuo[2]presented a polynomial–time classification algorithm. The same procedure was also carried out for the L( j, 1)-labelings. It was proved by Chang et al.[1]that for any tree of maximum degree∆,

∆ + j − 1 ≤ λj,1(T ) ≤ min{∆ + 2 j − 2, 2∆ + j − 2}.

Moreover, the lower bound and the upper bounds are both attainable. They also gave a polynomial–time algorithm for determiningλj,1(T ) of a tree T .

We now consider the lower bound on the above two cases for general graphs G and general positive integers j≥ k.

Proposition 1. For any positive integers j ≥ k and any graph G of maximum degree

∆ ≥ 1, we have λj,k(G) ≥ j + (∆ − 1)k. Moreover, if j > k and the equality holds, then

for anyλj,k-labeling of G each major vertex (i.e. a vertex of degree∆) must be labeled 0

(or j+ (∆ − 1)k) and its neighbors must be labeled j + ik (or ik) for i = 0, 1, . . . , ∆ − 1.

We call a graph G of maximum degree∆ ≥ 1 with λj,k(G) = j + (∆ − 1)k a λj,k

-minimal graph. It is clear that K1,∆ isλj,k-minimal for all∆ ≥ 1 and j ≥ k; see[5]. While determiningλj,k(G) for a graph is in general not an easy job, the aim of this paper is to study the structures ofλj,k-minimal graphs.

2. λλλjjj,k,k,k-minimal graphs

We first study the structures ofλj,k-minimal graphs.

Notice that whether a graph isλj,k-minimal or not depends on the values of j and k. As an example, for∆ ≥ 1 consider the double star D which is obtained from two stars

K1,∆by identifying a leaf of one star with a leaf of the other star.Fig. 1shows the double star D6.

ByProposition 1, if D isλj,k-minimal then one of the two major vertices is labeled by 0 and its neighbors by j, j +k, . . . , j +(∆−1)k, and the other is labeled by j +(∆−1)k and its neighbors by 0, k, . . . , (∆−1)k. Hence, the only common neighbor of the two major vertices is labeled by j+ i k = i k for some 0≤ i , i ≤ ∆ − 1. This is possible if and

only if j = ik for some integer 1 ≤ i ≤ ∆ − 1. In conclusion, Disλj,k-minimal if and only if j = ik for some integer 1 ≤ i ≤ ∆ − 1.

(3)

Fig. 1. The double star D6.

For some cases, λj,k-minimality does imply other λp,q-minimality. Consider the following result in[5].

Lemma 2. The following are true for any graph G.

(1) λcj,ck(G) = cλj,kfor all positive integers c and j ≥ k.

(2) λj,k(G) ≤ λp,q(G) for all positive integers j ≥ k and p ≥ q satisfying j ≤ p and

k≤ q.

We then have

Theorem 3. Suppose G is a graph of maximum degree∆ ≥ 1; and j ≥ k and p ≥ q are

positive integers satisfying p= q j/k. If G is λj,k-minimal, then G isλp,q-minimal.

Proof. According toLemma 2, we have

kλp,q(G) = λkp,kq ≤ λq j,qk = qλj,k= q( j + (∆ − 1)k).

Hence, λp,q(G) ≤ q j/k + (∆ − 1)q implying λp,q(G) ≤ p + (∆ − 1)q and so

λp,q(G) = p + (∆ − 1)q. This completes the proof of the theorem. 

As seen above, not only is determiningλj,k(G) hard, but also deciding whether G is

λj,k-minimal may be not easy as it depends on both the structure of G and the values of j and k. In the rest of the paper, we restrict our attention to graphs that areλp,q-minimal for all p≥ q.

We then have the following theorem.

Theorem 4. Suppose G is a graph of maximum degree∆ ≥ 1 and j ≥ k are positive

integers satisfying j≥ ∆k. Then the following statements are equivalent.

(1) G isλj,k-minimal.

(2) G has aλj,k-labeling g such that any vertexv in G, g(v) is of the form avj+ bvk

with av ∈ {0, 1} and bv ∈ {0, 1, . . . , ∆ − 1}. Moreover, the following statements hold.

(D1) If dG(u, v) = 1, then au= av. If au= 0 and av= 1, then bu≤ bv. (D2) If dG(u, v) = 2, then au= avand bu= bv.

(4)

(v) < f (u) and so f (v) ≤ f (u) − j ≤ (∆ − 1)k implying bv ≤ ∆ − 1. If such u does not exist, then we may assume that f(v) = 0. In any case, we have 0 ≤ bv≤ ∆ − 1.

Now, consider the function g on V(G) defined by g(v) = avj+ bvk for all verticesv

in G. We shall check that g is an L( j, k)-labeling of G as desired.

Suppose u and v are adjacent in G. As | f (u) − f (v)| ≥ j ≥ ∆k, it must be the case that au = av for otherwise| f (u) − f (v)| = |(auk+ ru) − (avk+ rv)| < ∆k is impossible. Without loss of generality, we may assume that au= 0 and av= 1. In this case,

f(u) < f (v) and so in fact f (u) ≤ f (v) − j implying bu≤ bv. Then g(u) ≤ g(v) + j. Next, suppose dG(u, v) = 2. Choose a vertex w adjacent to both u and v. Then, au= aw and aw = avimplying that au= av. So, we have k ≤ | f (u)− f (v)| = |buk+ru−bvk−rv| which implies bu= bv. Then|g(u) − g(v)| ≥ k.

(2) ⇒ (3). Define a labeling h : V (G) → {0, 1, 2, . . .} by h(v) = avp + bvq

for all vertices v in G. If dG(u, v) = 1, then (D1) implies |h(u) − h(v)| ≥ p; if

dG(u, v) = 2, then (D2) implies |h(u) − h(v)| ≥ q. Hence, h is an L(p, q)-labeling

of G with maxv∈V (G)h(v) ≤ p + (∆ − 1)q. Thus, G is λp,q-minimal.

(3) ⇒ (1). This follows by taking p = j and q = k. 

Corollary 5. If G is aλj,k-minimal graph of maximum degree∆ ≥ 1 and j ≥ ∆k, then

G is bipartite.

Proof. Notice that inTheorem 4, every vertex has a label of the form avj+ bvk uniquely.

According to condition (D1), au = avwhenever uv is an edge. Thus, G is a bipartite graph with a bipartition of V(G) into A = {v : av= 0} and B = {v : av= 1}. 

It is desirable to design an efficient algorithm to decide whether a graph isλp,q-minimal for all positive integers p≥ q. We are only able to do this for trees. This will be discussed in the next section.

3. λλλjjj,k,k,k-minimal trees

We now turn our attention to the case where the graph is a tree. For this case, an alternative statement ofTheorem 4is as follows.

For any positive integer∆, a ∆-sequence is a sequence (b0, b1, . . . , bm) of integers satisfying (S1) to (S4).

(S1) b0= 0.

(S2) 0≤ bi ≤ ∆ − 1 for all i.

(S3) bi ≥ bi−1and bi ≥ bi+1for all odd i . (S4) bi = bi+2 for all i .

(5)

We now define Tas the infinite tree (except that T1is K2) whose vertex set consisting all

∆-sequences and a vertex (b0, b1, . . . , bm) is adjacent to another vertex (c0, c1, . . . , cn) if and only if|m − n| = 1 and bi = cifor 0≤ i ≤ min{m, n}. Now, we have

Theorem 6. Suppose T is a tree of maximum degree∆ ≥ 1 and j ≥ k are positive

integers satisfying j≥ ∆k. Then, T is λj,k-minimal if and only if T is a subtree of T∆.

Proof. (⇒) ByTheorem 4, T has aλj,k-labeling g such that any vertexv is labeled by

g(v) = avj+ bvk where av ∈ {0, 1} and bv ∈ {0, 1, . . . , ∆ − 1} satisfying conditions

(D1) and (D2) inTheorem 4. We may view T as a tree rooted at a major vertexv0. Without loss of generality, we may assume g(v0) = 0 by theProposition 1. For any vertexv in T , there is a unique path fromv0tov, denoted by P : v0, v1, . . . , vm. According to (D1) and (D2),(bv0, bv1, . . . , bvm) is a ∆-sequence. Then there is a one-to-one function from V (T ) to V(T) mapping v to the corresponding ∆-sequence (bv0, bv1, . . . , bvm). It is clear that this mapping is edge-preserving. Hence T is a subtree of T.

(⇐) According toProposition 1,λj,k(T ) ≥ j + (∆ − 1)k. Since T is a subtree of T∆, it suffices to give an L( j, k)-labeling of Twhose span is j+ (∆ − 1)k. Defin f as follows: for any vertex(b0, b1, . . . , bm) in T∆, let

f(b0, b1, . . . , bm) = 

j+ bmk, if m is odd;

bmk, if m is even.

By conditions (S1) to (S4) for the definition of a∆-sequence, it is straightforward to check that f is an L( j, k)-labeling of T. 

Note that we may determine whether a tree T of maximum degree ∆ ≥ 1 is λp,q -minimal for all positive integers p≥ q or not by just checking whether it is λ∆,1-minimal. This can be done by findingλ∆,1(T ) using the algorithm described in[1]. We are expecting a simpler algorithm from just checking whether T is a subtree of T.

Acknowledgements

We thank the referee for many constructive suggestions. This paper was supported in part by the National Science Council under grant NSC89-2115-M009-037.

References

[1] G.J. Chang, W.-T. Ke, D. Kuo, D.D.-F. Liu, R.K. Yeh, On L(d, 1)-labelings of graphs, Discrete Math. 220 (2000) 57–66.

[2] G.J. Chang, D. Kuo, The L(2, 1)-labeling on graphs, SIAM J. Discrete Math. 9 (1996) 309–316. [3] G.J. Chang, S.-C. Liaw, The L(2, 1)-labeling problem on ditrees, Ars Combin. (accepted).

[4] J. Georges, D.W. Mauro, On the criticality of graphs labeled with a condition at distance two, Congr. Numer. 101 (1994) 33–49.

[5] J. Georges, D.W. Mauro, Generalized vertex labelings with a condition at distance two, Congr. Numer. 109 (1995) 141–159.

[6] J. Georges, D.W. Mauro, On the size of graphs labeled with a condition at distance two, J. Graph Theory 22 (1996) 47–57.

[7] J. Georges, D.W. Mauro, Some results onλkj-numbers of the products of complete graphs, Congr. Numer. 140 (1999) 141–160.

(6)

domination, IEEE Trans. Circuits Syst. I 47 (2000) 1531–1534.

[13] P.K. Jha, A. Narayanan, P. Sood, K. Sundaram, V. Sunder, On L(2, 1)-labeling of the Cartesian product of a cycle and a path, Ars Combin. 55 (2000) 81–89.

[14] D.D.-F. Liu, R.K. Yeh, On distance-two labelings of graphs, Ars Combin. 47 (1997) 13–22.

[15] J. Ne˘set˘ril, Aspects of structural combinatorics (graph homomorphisms and their use), Taiwanese J. Math. 3 (1999) 381–423.

[16] F.S. Roberts, Private communication through J. Griggs (1988).

[17] D. Sakai, Labeling chordal graphs with a condition at distance two, SIAM J. Discrete Math. 7 (1994) 133–140.

[18] M. Whittlesey, J. Georges, D.W. Mauro, On theλ-number of Qn and related graphs, SIAM J. Discrete Math. 8 (1995) 499–506.

[19] K.-F. Wu, R.K. Yeh, Labeling graphs with the circular difference, Taiwanese J. Math. 4 (2000) 397–405. [20] R.K. Yeh, The edge span of distance two labelings of graphs, Taiwanese J. Math. 4 (2000) 675–683.

數據

Fig. 1. The double star D 6 .

參考文獻

相關文件

• When light is refracted into two rays each polarized with the vibration directions.. oriented at right angles to one another, and traveling at

He(She) has a round face and two small eyes.. Are there

Curriculum planning: Teachers may introduce how to use short division to find the greatest common divisor and the least common multiple of two numbers at Learning Objective 1.4

We must assume, further, that between a nucleon and an anti-nucleon strong attractive forces exist, capable of binding the two particles together.. *Now at the Institute for

Quantum Hall Effect in Black Phosphorus 2DEG.. Likai Li

In this paper, we extend this class of merit functions to the second-order cone complementarity problem (SOCCP) and show analogous properties as in NCP and SDCP cases.. In addition,

Midpoint break loops are useful for situations where the commands in the loop must be executed at least once, but where the decision to make an early termination is based on

¾ PCS systems can connected to Public Switched Telephone Network (PSTN)6. ¾ Goal of PCS:enabling communications with a person at anytime, at any place and in any