1 2

Discussiones Mathematicae Graph Theory xx (xxxx ) 117

3 4

L(2, 1)-LABELINGS OF SOME FAMILIES OF ORIENTED PLANAR GRAPHS

Sagnik Sen1
Univ. Bordeaux, LaBRI, UMR5800, F-33400 Talence, France. CNRS, LaBRI, UMR5800, F-33400 Talence, France. e-mail: [email protected]

6 7 8

9 10 11 12 13 14 15

Abstract In this paper we determine, or give lower and upper bounds on, the 2-dipath and oriented L(2, 1)-span of the family of planar graphs, planar graphs with girth 5, 11, 16, partial k -trees, outerplanar graphs and cacti. Keywords: 2-dipath L(2, 1)-labeling, oriented L(2, 1)-labeling, homomorphism, planar graph, girth, partial k -tree, outerplanar graph, cactus. 2010 Mathematics Subject Classication: 05C78.

16

1.

Introduction

17 18 19 20 21 22 23 24 25 26 27 28 29

To distinguish close and very close transmitters in a wireless communication system, Griggs and Yeh [6] proposed a variation of the Frequency Assignment Problem (or simply FAP) by introducing the L(2, 1)-labeling which was generalized by Georges and Mauro [4] as follows. For any two positive integers p and q , a k-L(p, q )-labeling of a graph G is a mapping from the vertex set V (G) to the set {0, 1, ...., k } such that - | (u) (v ) | p if u and v are at distance 1 in G, - | (u) (v ) | q if u and v are at distance 2 in G. The L(p, q )-span p,q (G) of a graph G is dened as min{k | G has a k -L(p, q )labeling }. For a family F of graphs, p,q (F ) = max{p,q (H ) | H F}. A common feature of graph theoretic models for FAP is that communication is assumed to be possible in both directions (duplex) between two radio transmitters and, therefore, these models are based on undirected graphs. But in reality,
1

This work was supported by ANR GRATEL project ANR-09-blan-0373-01.

S. Sen

30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59

to model FAP on directed or oriented graphs could be interesting as pointed by Aardal et al [1] in their survey. An oriented graph is a directed graph with no cycle of length 1 or 2. By replacing each edge of a simple graph G with an arc (ordered pair of vertices) we obtain an oriented graph G; G is an orientation of G and G is the underlying graph of G. We denote by V (G) and A(G) respectively the set of vertices and arcs of G. Similarly, V (G) and E (G) denote respectively the set of vertices and edges of G. An arc (x, y ) (where x and y are vertices) is denoted by xy . A path obtained by two consiqutive arcs xy and yz is called a 2-dipath. Every undirected graph is a simple graph and every directed graph is an oriented graph in this paper unless otherwise stated. There are two dierent oriented versions of L(p, q )-labeling, namely 2-dipath L(p, q )-labeling, introduced by Chang et al [3], and oriented L(p, q )-labeling, introduced by Gon calves, Raspaud and Shalu [5]. A 2-dipath k-L(p, q )-labeling of an oriented graph G is a mapping from the vertex set V (G) to the set {0, 1, ...., k } such that - | (u) (v ) | p if u and v are adjacent in G, - | (u) (v ) | q if u and v are connected by a 2-dipath in G. The 2-dipath span p,q (G) of an oriented graph G is dened as min{k | G has a 2-dipath k -L(p, q )-labeling }. The 2-dipath span p,q (G) of an undirected graph G is dened as max{p,q (G) | G is an orientation of G}. The 2-dipath span p,q (F ) of a family F of (oriented or undirected) graphs is dened as max{p,q (H ) | H F}. An oriented k -L(p, q )-labeling of an oriented graph G is a mapping from the vertex set V (G) to the set {0, 1, ...., k } such that - is a 2-dipath k -L(p, q )-labeling of G, - if xy and uv are two arcs in G then, (x) = (v ) implies (y ) = (u). o o The oriented spans o p,q (G), p,q (G) and p,q (F ) are dened similarly as 2dipath spans. From the denitions, the following are immediate: Lemma 1. For every (undirected or oriented) graph G and every p, q > 0, p,q (G) o p,q (G). The additional condition in oriented L(p, q )-labeling ensures that any oriented L(p, q )-labeling is an oriented coloring [11]. An oriented k -coloring of an oriented graph G is a mapping f from the vertex set V (G) to the set {0, 1, ...., k 1} such that, - f (u) = f (v ) whenever u and v are adjacent, - if xy and uv are two arcs in G then, f (x) = f (v ) implies f (y ) = f (u). The oriented chromatic number (G) of an oriented graph G is dened as min{k | G has an oriented k -coloring}. The oriented chromatic number (G) of

60 61

62 63 64 65 66 67 68 69

L(2, 1)-labelings of some families of oriented planar graphs

70 71 72 73 74 75 76 77 78 79 80 81

an undirected graph G is dened as max{(G) | G is an orientation of G}. The oriented chromatic number (F ) of a family F of (oriented or undirected) graphs is dened as max{(H ) | H F}. Note that any oriented k -L(p, q )-labeling is an oriented (k + 1)-coloring but a 2-dipath k -L(p, q )-labeling is not necessarily an oriented (k + 1)-coloring. The most frequently studied L(p, q )-labeling, other than the ones that correspond to chromatic numbers, is for (p, q ) = (2, 1) (both undirected and oriented versions). In this paper, we mainly focus on studying 2-dipath and oriented L(2, 1)-span of some families of planar graphs. For the family P of planar graphs and for the family Pg of planar graphs with girth at least g , where the girth of a graph is the size of its smallest cycle, for g = 5, 11 and 16, we will prove the following result in section 3. Theorem 2. (a) 22 2,1 (P ) o (P ) 83. (b) 6 2,1 (P5 ) o (P5 ) 22. (c) 4 2,1 (P11 ) o (P11 ) 10. (d) 4 2,1 (P16 ) o (P16 ) 7. Theorem 2(b) disproves the conjecture 2,1 (P5 ) 5 proposed by Calamoneri and Sinaimeri [2] and Theorem 2(c,d) improve the previous bounds 2,1 (P11 ) 12 and 2,1 (P16 ) 8 given by the same authors [2]. To prove the lower bound in Theorem 2(a), we will construct an outerplanar graph O with 2-dipath L(2, 1)span at least 9 (section 5).For the family O of outerplanar graphs, we prove in section 5 the following: Theorem 3. 9 2,1 (O) o (O) 10. As we were not able to provide exact results for the family of outerplanar graphs, we also consider a planar superfamily and a planar subfamily of it, namely the family T2 of partial 2-trees and the family C of cacti. For both these families we managed to give exact results. In fact, in section 4 we prove the following general result for the family Tk of partial k -trees: Theorem 4. (a) 2,1 (T2 ) = o (T2 ) = 10. (b) 2,1 (T3 ) o (T3 ) 22. (c) 2,1 (Tk ) o (Tk ) (k + 1)(2k + 1) 2. In [2] Calamoneri and Sinaimeri proved that 6 2,1 (C ) 8. We improve this result as follows (proof in section 6): Theorem 5. 2,1 (C ) = o p,q (C ) = 7.

82 83 84 85

86 87 88 89 90 91

92

93 94 95 96 97

98 99 100

101 102

103

4 7 6 0 1

S. Sen

5 4 3

Figure 1. B is a 4-nice graph.

104 105 106 107 108

The precise denition of these families are given in the begining of their respective sections. In section 2, we mainly dene, state and prove some results which we will use for the main proofs in the following sections. In particular, we prove a general upper bound for the 2-dipath and oriented L(p, q )-span of multipartite graphs, which in some cases is tight.

109

2.

Preliminaries

110 111 112 113 114 115 116 117 118 119

The set of all adjacent vertices of a vertex v in a graph is called its set of neighbors and is denoted by N (v ). For oriented graphs, if there is an arc from u to v , then u is an in-neighbor of v and v is an out-neighbor of u. The sets of all in-neighbors and out-neighbors of v are denoted by N (v ) and N + (v ) respectively. An oriented clique or simply oclique is an oriented graph whose any two distinct vertices are either adjacent or connected by a 2-dipath. Ocliques are therefore precisely those oriented graphs G for which (G) =| V (G) |. A homomorphism f of an oriented graph G to an oriented graph H is a mapping f : V (G) V (H ) such that xy A(G) implies f (x)f (y ) A(H ). From these denitions, we easily get the following: Lemma 6. If there is a homomorphism f : G H then p,q (G) p,q (H ) and o o p,q (G) p,q (H ), for every p, q > 0. o In particular, G H implies p,q (G) p,q (H ) and o p,q (G) p,q (H ), for all p, q > 0. Now we prove a general upper bound on oriented L(p, q )-span of multipartite graphs. Theorem 7. For every k -partite oriented graph G, where k 3, we have

120 121 122 123

124 125

126

L(2, 1)-labelings of some families of oriented planar graphs

127 128 129

p,q (G) o p,q (G) | V (G) | q + k (max(p, q ) q ) max(p, q ). In particular for p q , both the equalities hold if G is a complete k -partite oclique. Proof. Let, K = Kn1 ,n2 ,...,nk be the complete k -partite graph with the parts being V1 , V2 , ...Vk with | Vi |= ni for all i = 1, 2, ..., k . Also, let the vertices of Vi be denoted by {vi1 , vi2 , ..., vini }. Let K be any orientation of K . Now, consider the labeling L of K given by L(vij ) = ( t<i (nt 1)q ) + (j 1)q + (i 1)max(p, q ), for i = 1, 2, ..., k and j = 1, 2, ..., ni . For any i, vir and vis (r = s) cannot be connected by an arc but can be connected by a 2-dipath. While for any vir and vjs , i = j , can be connected by either an arc or a 2-dipath. Then we have, | L(vir ) L(vis ) | 1)q ) + (r 1)q + (i 1)max(p, q )] [( t<i (nt 1)q ) + (s 1)q + (i 1)max(p, q )] = |[( t<i (nt 1)q ) ( t<i (nt 1)q )]+ [(r 1)q (s 1)q ] + [(i 1)max(p, q ) (i 1)max(p, q )] = |(r s)q | q , for r = s = |[(
t<i (nt

130 131 132 133 134 135 136 137 138 139

140 141 142 143 144

145

and, | L(vir ) L(vjs ) | = |[( 1)q ) + (r 1)q + (i 1)max(p, q )] [( t<j (nt 1)q ) + (s 1)q + (j 1)max(p, q )]| = |( j<t<i (nt 1)q ) + (nj 1)q + (r 1)q (s 1)q + (i j )max(p, q )| (without loss of generality, we assume i > j ) = |( j<t<i (nt 1)q ) + (nj s)q + (r 1)q + (i j )max(p, q )| (nj s as vjs Vj ) max(p, q ).
t<i (nt

146 147 148 149 150 151 152

153 154 155 156 157 158 159 160 161 162 163

As all vertices have dierent labels, L is an oriented coloring of K . Hence we have, k 1 o p,q (K ) p,q (K ) t=1 (nt 1)q + (nk 1)q + (k 1)max(p, q ) = =| V (K ) | q + k (max(p, q ) q ) max(p, q ). Now as any oriented k -partite graph G is a subgraph of some orientation of the complete k -partite graph K , using Lemma 1 and Lemma 6 the theorem follows. In particular, if K is an oclique, then any two vertices are at distance at most 2. Moreover, if K is also an orientation of the complete k -partite graph, then any two vertices from dierent parts, are adjacent. Hence both the equalities hold for p q.

S. Sen

164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184

For any prime p 3(mod 4) and for any positive integer n the Paley tournament Pq of order q = pn is the oriented graph with set of vertices {0, 1, 2, ...., q 1} and set of arcs {xy | y x (mod p) is a non-zero square}. As 1 (mod p) is a not a square , either (x y ) or (y x) (but not both) is a square for all x, y Fq . Hence Pq is a tournament. The Tromp graph [11] T2q+2 of order (2q + 2) is the oriented graph with set of vertices V (T2q+2 ) = {0, 1, ..., (q 1)} {0 , 1 , ..., (q 1) } {v, v } and set of arcs A(T2q+2 ) = {ij, i j , j i, ji | i, j {0, 1, ..., (q 1)} and (j i) (mod p) is a non-zero square } {iv, vi , i v , v i | i {0, 1, ..., (q 1)}}. Intuitively, in T2q+2 there are two vertices v, v such that N + (v ) = N (v ) and N + (v ) = N (v ) with each of the sets N + (v ) and N (v ) inducing a Paley tournament Pq . Also, if ij is an arc in the Pq induced by N + (v ) and i j is the corresponding arc of the Pq induced by N (v ), then we also have the arcs ji and j i. Note that, T2q+2 is a complete (q + 1)-partite oclique with all parts of size two. For further details about this graph, see Marshalls paper [8]. For any positive integer k the Zielonka graph [11] Zk of order k 2k1 is the oriented graph , with set of vertices V (Zk ) = i=1,2,...,k Si , where Si = {x = (x1 , ..., xk )|xj {0, 1} for j = i and xi = } and set of arcs A(Zk ) = {xy | x = (x1 , ..., xk ) Si , y = (y 1 , ..., y k ) Sj and either xj = y i and i < j or xj = y i and i > j }. Note that, Zk is a complete k -partite oclique with all parts of size 2k1 . So, by Theorem 7 we have the following: Corollary 8. 2,1 (T2q+2 ) = o (T2q+2 ) = 3q + 1. Corollary 9. 2,1 (Zk ) = o (Zk ) = k (2k1 + 1) 2. In this paper, we shall use the following notion from [7]. A pattern Q of length k is a word Q = q0 q1 ...qk1 with qi {+, } for every i, 0 i k 1. A Q-walk in a digraph G is a walk P = x0 x1 .....xk such that for every i, 0 i k 1, xi xi+1 A(G) if qi = + and xi+1 xi A(G) otherwise. For X V (G) we denote by NQ (X ) the set of all vertices y such that there exists a Q-walk going from some vertex x X to y . We then say that a digraph G is k -nice if for every pattern Q of length k and for every vertex x V (G) we have NQ ({x}) = V (G). In other words, a digraph is k -nice if for all pairs of vertices x, y (allowing x = y ) there is a k -walk from x to y for each of the 2k possible oriented patterns. Observe that if a digraph G is k -nice for some k , then it is k -nice for every k k . For example, the graph B (Fig 1) is a 4-nice graph. The girth of a graph is the length of its shortest cycle. We denote the family of planar graphs by P and the family of planar graphs with girth at least g by Pg . Now we state a theorem from [7].

185

186

187 188 189 190 191 192 193 194 195 196 197 198 199 200 201

L(2, 1)-labelings of some families of oriented planar graphs v u a2 a1 x y b1 b2

Figure 2. F is an oriented planar graph with girth 5.

202 203 204

Theorem 10 Hell et al. 1997. Let Nk be a k -nice oriented graph, k 3. Every oriented graph whose underlying graph is in P5k4 admits a homomorphism to Nk .

205

3.

Planar graphs

206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228

A planar graph is a graph that can be drawn in the plane in such a way that no two edges cross each other. Now we prove Theorem 2. Proof of Theorem 2 (a) Raspaud and Sopena [10] showed that every oriented planar graph admits a homomorphism to the Zielonka graph Z5 . Hence, using Lemma 6 and Corollary 9, we get the upper bound. For proving the lower bound, assume O is an outerplanar graph with 2dipath span at least 9. Now, take a 2-dipath Q along with six disjoint copies of O . Then, connect (dierent) two copies of O with each of the vertices of Q by adding arcs. We choose orientations of the new arcs in such a way that, for each vertex v V (Q), the graph induced by N (v ) contains O as a subgraph for = +, . Name this graph, R. Notice that R is a planar graph. Now let f be a 2-dipath k -L(2, 1)-labeling of R for some k . Then there will be a vertex v V (Q) such that f (v ) = 0, k . This means f (t) / {f (v ) 1, f (v ), f (v )+ 1} for t N (v ). Also, we know that f (x) = f (y ) for x N + (v ) and y N (v ). Moreover, as O is a subgraph of both N + (v ) and N (v ) we need at least 20 labels other than {f (v ) 1, f (v ), f (v ) + 1} to label N (v ). Hence, k 22. We will construct such an outerplanar graph O in the proof of Theorem 3 (5). That will complete the proof of Theorem 2(a). (b) We know that every planar graph of girth at least 5 admits a homomorphism to the Tromp graph T16 [9]. Then, using Lemma 6 and Corollary 8, we get the upper bound.

S. Sen

x4 x6 x5 x1 x2 x3

Figure 3. E is an oriented planar graph with girth 5.

229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255

To prove the lower bound, we rst show that it is impossible to have a 2dipath 5-L(2, 1)-labeling f of the graph F , depicted in Figure 2, with {f (x), f (y )} = {3, 5}. Notice that, if f (x) = 3 and f (y ) = 5 then f (a1 ) {0, 1} and f (b1 ) {0, 1, 2}. This implies f (u) = 4. Similarly, we have f (v ) = 4 which is not possible as u, v are adjacent. The case f (x) = 5 and f (y ) = 3 is similar. The oriented planar graph E , depicted in Figure 3, has girth 5. Moreover, the vertices {x1 , x2 , x3 , x4 , x5 , x6 } will get pairwise dierent labels for any 2dipath L(2, 1)-labeling since they are pairwise connected by a 2-dipath. Consider a 2-dipath 5-L(2, 1)-labeling g of E such that g (x6 ) = 0. Then we have {g (x1 ), g (x2 ), g (x3 ), g (x4 ), g (x5 )} = {1, 2, 3, 4, 5}. Hence there exists an arc wz A(G[{x1 , x2 , x3 , x4 , x5 }]) such that {f (w), f (z )} = {3, 5}. Now on each of the ve vertices x {x1 , x2 , x3 , x4 , x5 } of E , we glue a copy of E by identifying x with the vertex x6 of the copy. Call this graph G. Note that G is a planar graph with girth 5 such that, for any 2-dipath 5L(2, 1)-labeling of G, there is wz A(G) with { (w), (z )} = {3, 5}. We can then construct a new graph H by identifying each arc ab of G with the arc xy of F . Clearly, H is also a planar graph with girth 5 which does not have any 2-dipath 5-L(2, 1)-labeling. Hence we get the lower bound which completes the proof. (c), (d) It is easy to observe that the directed path of length 5 has 2-dipath L(2, 1)-span 4. Now, given any girth g , there is an oriented planar graph with girth g containing the directed path of length 5 as a subgraph. Also, one can check that the Tromp graph T8 is 3-nice, and that the graph B (Fig 1) is 4-nice (we have veried both using computer). Then, using Theorem 10, Lemma 6, Lemma 1 and Corollary 8, we have the results.

L(2, 1)-labelings of some families of oriented planar graphs 4. Partial k -trees

256

257 258 259 260 261 262 263 264 265 266 267 268 269

A k -tree is a graph obtained from the complete graph Kk on k vertices by adding zero or more vertices, one by one, in such a way that each newly added vertex has exactly k neighbors that form a clique. A subgraph of a k -tree is a partial k -tree We denote the family of partial k -trees by Tk . Partial k -trees have been extensively studied in the last past years, since they often lead to polynomial algorithms for problems which are known to be NP-complete in the general case [14]. The notion of a 1-tree obviously corresponds to the usual notion of a tree. The family of outerplanar graphs is strictly contained in the family of partial 2-trees. It is easy to see that every partial 2-tree is a planar graph. The following Lemmas will be useful for proving Theorem 4. To prove the following lemma we use the same technique as the one used to prove that every oriented outerplanar graph has oriented chromatic number at most 7 in [12]. Lemma 11. Every oriented partial 2-tree D admits a homomorphism to the Tromp graph T8 . Proof. It is possible to check that for every u, v V (T8 ) and every , {+, }, there exists w N (u) N (v ). Let G be a minimal (with respect to the number of vertices) counterexample to the lemma. Without loss of generality, we may assume that G is a 2-tree. Since G is a 2-tree, G must have a vertex x of degree 2. Let N (x) = {x1 , x2 }. Now, by removing the vertex x from G and adding an arc between x1 and x2 (if there was not already one), we get a 2-tree that admits a homomorphism to T8 (because of the minimality of G). Using the property of T8 stated in the begining of the proof, clearly this homomorphism can be extended to a homomorphism of G to T8 , a contradiction. Lemma 12. There exists an oriented 2-tree D13 for which 2,1 (D13 ) 10. Proof. First, we will describe a family of oriented 2-trees by induction. We start with the oriented 2-tree D1 (Fig: 4). By induction, we construct a graph Di+1 by gluing D1 on each arc of Di by identifying that arc with the arc xy of D1 . Note that every so obtained graph Di is a 2-tree. Assume that f is a 9-L(2, 1)-labeling of D13 . Step 0: Notice that, in each copy of D1 , all the vertices should get dierent labels and for any vertex v N (x) N (y ) we have, |f (t) f (v )| 2 for t = x, y . Step 1: If we restrict f to D1 then there is a vertex v1 N (x) N (y ) such that, f (v1 ) / {0, 9}. Similarly, if we restrict f to D2 , we can nd a vertex v2 N (x) N (v1 ) such that, f (v2 ) / {0, 9}.

270 271

272 273 274 275 276 277 278 279 280 281

282

283 284 285 286 287 288 289 290 291 292

10

S. Sen

Figure 4. The oriented 2-tree D1

293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324

Step 2: Now, if we restrict f to D3 , we can nd a v3 N (v1 ) N (v2 ) such that, f (v3 ) / {0, 9}. So we have {f (v1 ), f (v2 ), f (v3 )} {1, 2, 3, 4, 5, 6, 7, 8} with no two of {f (v1 ), f (v2 ), f (v3 )} being consecutive numbers, since {v1 , v2 , v3 } are pairwise adjacent vertices. Hence there exists i, j {1, 2, 3} such that {f (vi ) 1, f (vi ), f (vi ) + 1} {f (vj ) 1, f (vj ), f (vj ) + 1} = and f (vi ) < f (vj ). Step 3: In D4 there exists v N (vi ) N (vj ) for all , {+, }. Notice that the vertices {v++ , v+ , v+ , v } will be labeled by the four remaining labels dierent from the labels {f (vi ) 1, f (vi ), f (vi ) + 1, f (vj ) 1, f (vj ), f (vj ) + 1}. Step 4: Now we want to show that there is a vertex in D5 that receives label 1 or 8. If f (t) {1, 8} for some t {vi , vj , v++ , v+ , v+ , v }, we are done. If not, then we can conclude that f (vi ) = 2, f (vj ) = 7 since any other possible choice of labels (other than 1 or 8) for vi , vj will force at least one of the labels among {f (v++ ), f (v+ ), f (v+ ), f (v )} to be equal to 1 or 8. This will imply {f (v++ ), f (v+ ), f (v+ ), f (v )} = {0, 4, 5, 9}. Choose v4 from the set {v++ , v+ , v+ , v } such that, f (v4 ) = 5. Then in D5 , there is a vertex v5 N (vi ) N (v4 ) with f (v5 ) = 8. Hence in D5 , there exists a vertex v6 with f (v6 ) {1, 8}. Step 5: Now we want to show that there is a vertex in D7 that receives label 1. If f (v6 ) = 1, we are done. If not, then f (v6 ) = 8. This implies that, in D6 , there exists t N (v6 ) such that f (t) {1, 4, 5}, since we need to use at least ve distinct labels from {0, 1, 2, 3, 4, 5, 6} to label all vertices of N (v6 ). If f (t) = 1, we are done. Otherwise, in D7 , we can nd some s N (v6 ) N (t) such that f (s) = 1. Hence in D7 we can nd a vertex a with f (a) = 1. Step 6: Now we want to show that in D9 there is a vertex b N (a) with f (b) = 8. Now, in D8 , there are at least ve vertices in N (a) which receive pairwise dierent labels. Therefore, for some t N (a), we will have f (t) {8, 4, 5}. If f (t) = 8, we are done. Otherwise, in D9 , we can nd s N (a) N (t) with f (s) = 8. Hence, in D9 , there is a pair of adjacent vertices a and b with f (a) = 1 and

L(2, 1)-labelings of some families of oriented planar graphs

11

325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362

f (b) = 8. Step 7: Therefore, in D10 , there will be a copy of D1 with vertices {a, b} corresponding to the vertices {x, y } of D1 (as in Fig 4). Step 8: Now notice that, in D12 , there are ui N (a) such that f (ui ) = i for all i {3, 4, ..., 9}. Hence, in D13 , there are u N (a) N (ui ) for all i , {+, }. Step 9: Note that it is not possible to have p N + (a) and q N (a) with f (p) = f (q ). Hence the function Fa (i) = if t N (a) and f (t) = i for i {3, 4, ..., 9} and {+, }, is well dened. Intuitively, the function Fa is a function indicating whether a label is used for an in-neighbor of a or for an out-neighbor of a. + Step 10: Note that for each i {3, 4, ..., 9}, Fa (f (u++ i )) = Fa (f (ui )) = + + and Fa (f (ui )) = Fa (f (ui )) = . Also, notice that {f (u i )|, {+, }} = {3, 4, ..., 9} \ {i 1, i, i + 1} for each i {4, 5, 6, 7, 8}. We will use the two above observations repeatedly in the following. Step 11: Let {, } = {+, }. Without loss of generality, assume that Fa (3) = . Claim: Fa (6) = . Proof of the claim: If possible, let Fa (6) = . Now {f (u 8 )|, {+, }} = {3, 4, 5, 6}. So two of Fa (3), Fa (4), Fa (5), Fa (6) will be + and the other two will be . But we already have Fa (3) = and Fa (6) = . Hence, {Fa (4), Fa (5)} = {, }. Similarly, we have {f (u 7 )|, {+, }} = {3, 4, 5, 9}. This will force Fa (9) = . After that we have {f (u 4 )|, {+, }} = {6, 7, 8, 9} which forces Fa (7) = Fa (8) = . Now we also have {f (u 5 )|, {+, }} = {3, 7, 8, 9}. But Fa (3) = Fa (7) = Fa (8) = , a contradiction. Hence, Fa (6) = . Step 12: Now {f (u 8 )|, {+, }} = {3, 4, 5, 6} implies Fa (4) = Fa (5) = . Similarly, {f (u7 )|, {+, }} = {3, 4, 5, 9} implies Fa (9) = . Lastly {f (u 4 )|, {+, }} = {6, 7, 8, 9} implies Fa (7) = Fa (8) = . Hence, we got the full description of Fa (depending on the value of ). Step 13: Similarly, we can dene a function Fb (one can imitate the previous steps, or just use symmetry). As f (a) = 1, f (b) = 8 and Fa (8) = , we have Fb (1) = . Now, by symmetry we get Fb (1) = Fb (2) = Fb (4) = Fb (5) = and Fb (0) = Fb (3) = Fb (6) = . Step 14: Therefore, Fa (l) = Fb (l) for all such l on which both the functions ++ are dened. But we have Fa (f (u++ 8 )) = Fb (f (u8 )) = +. This is a contradiction. Hence, we are done. We are now able to prove Theorem 4.

363 364

12 a1 a2 a3 a4 a5 a6 a7 a8 b1 b2 b3 b4 b5 b6 b7 b8

S. Sen

y7 x8 y6 y5 y4 y3 y2 y1 x1 x2

x3 x4 x5 x6 y8 x7

Figure 5. The oriented outerplanar graph O

365 366 367 368 369 370

Proof of Theorem 4 (a) The proof follows from Lemma 11, Lemma 12, Lemma 6 and Corollary 8. (b), (c) From [11] we know that any partial 3-tree admits a homomorphism to the tromp graph T16 and that any partial k -tree admits a homomorphism to the Zielonka graph Zk+1 . Hence the proof follows using Lemma 6 and Corollaries 8 and 9.

371

5.

Outerplanar graphs

372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389

A graph G is outerplanar if the graph formed from G by adding a new vertex, with edges connecting it to all the other vertices, is a planar graph. We denote the family of outerplanar graphs by O. Now we prove Theorem 3. Proof of Theorem 3 Every outerplanar graph is also a partial 2-tree. So, the upper bound follows from Theorem 4. To prove the lower bound, we will construct an oriented outerplanar graph O with 2,1 (O ) 9. This will complete the proof. First, we show that the outerplanar graph O (Figure 5) has no 2-dipath 8-L(2, 1)-labeling if v gets label 1. Let f be a 2-dipath 8-L(2, 1)-labeling of O such that f (v ) = 1. This implies f (t) / {0, 1, 2} for t {x1 , x2 , ..., x8 , y1 , ..., y8 } and f (xi ) = f (yj ) for any i, j = 1, 2, ..., 8. Clearly, we need at least 3 distinct labels for each of the sets {xi | i = 1, ..., 8} and {yi | i = 1, ..., 8}. Also, if we use exactly 3 labels for either of these sets, then those 3 labels should have pairwise dierence at least 2. To satisfy the above conditions, by symmetry, we may assume without loss of

L(2, 1)-labelings of some families of oriented planar graphs

13

390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427

generality that we use labels {3, 5, 7} for {x1 , ..., x8 } and {4, 6, 8} for {y1 , ..., y8 }. Now, with these assumptions, the following conditions are forced: (a) f (bi ) / {f (x1 ) 1, f (x1 ), f (x1 ) + 1} for i = 1, 2, ..., 8. (b) f (bi ) = f (bj ) for i {1, 2, 3, 4} and j {5, 6, 7, 8}. (c) f (bi ) / {1, f (x2 ), f (y1 )} for i = 1, 2, 3, 4. (d) f (xi ) = f (xi+3 ) for all i = 1, 2, 3. (e) we need at least 3 distinct labels for either of the sets {b1 , ..., b4 } and {b5 , ..., b8 } for i = 1, ..., 8. Also, if we use exactly 3 labels for either of these sets, then those 3 labels should have mutual dierence at least 2. We have three cases to consider. Case 1: If f (x1 ) = 7, then f (y1 ) = 4 and f (x2 ) = 3 or 5. Then, {f (b1 ), f (b2 ), f (b3 ), f (b4 )} = {0, 2, 5} (by (a), (c), (e) ). This implies {f (b5 ), f (b6 ), f (b7 ), f (b8 )} = {1, 3, 4} (by (a), (b) ) which contradicts (e). Case 2: If f (x1 ) = 5, then f (y1 ) = 8 and f (x2 ) = 3 or 7. Then, {f (b1 ), f (b2 ), f (b3 ), f (b4 )} = {0, 2, 7} (by (a), (c), (e) ). Hence f (x2 ) = 3. This implies f (x3 ) = 7. Therefore, f (x6 ) = 7 (by (d) ). Now, the only possibility is to have f (y8 ) = 4 which will force f (x7 ) = 7 since f (x7 ) {3, 5, 7}. But x6 and x7 cannot have same labels since they are connected by a 2-dipath through y8 . This is a contradiction. Case 3: If f (x1 ) = 3, then f (y1 ) = 6 or 8 and f (x2 ) = 5 or 7. Then, {f (b1 ), f (b2 ), f (b3 ), f (b4 )} = {0, 5, 8} (by (a), (c), (e) ). This implies {f (b5 ), f (b6 ), f (b7 ), f (b8 )} = {1, 6, 7} (by (a), (b) ) which contradicts (e). Hence, we do not have a 8-L(2, 1)-labeling f of O such that f (v ) = 1. By symmetry, we can say that we do not have a 8-L(2, 1)-labeling f of O such that f (v ) = 7. Now dene S = V (O) \ {x2 , x7 , x8 , y2 , y7 , y8 } and let G = O[S ]. Notice that if we try to 2-dipath 8-L(2, 1)-label G, then we need to use 3 dierent labels for the vertices v, x1 and y1 . One of these three vertices should have a label l / {0, 8}. To label the neighbors of that vertex, we clearly need at least 6 labels other than l 1, l and l + 1. So, we have to use all the remaining 6 labels and whatever the value of l may be, we necessarily use label 1 or 7 to 2-dipath 8-L(2, 1)-label G. Now, we construct a new graph O by gluing a copy of O on each vertex of G by identifying that vertex of G with the vertex v of O. Note that O is an outerplanar graph that cannot be 2-dipath 8-L(2, 1)labelled, which proves the theorem.

14
+ v3 + v1

S. Sen u+ 1 z4
+ v2 v5

u+ 3 u+ 2 u+ 5

y4 y2 u 5 x u u 4 y1 y3 u 1 u+ 4 u 2 u 3

z2
+ v5

v
v4 v2 + v4

z1
v1

v3

z3

Figure 6. The oriented cactus H .

428

6.

Cacti

429 430 431

A cactus is a connected graph in which any two cycles can have at most one vertex in common. We denote the family of cacti by C . Now we prove Theorem 5. The following Lemmas will be useful for proving Theorem 5. Lemma 13. There exists an oriented cactus C with 2,1 (C ) 7. Proof. Let H be the oriented cactus depicted in Figure 6. We rst show that there is no 6-L(2, 1)-labeling f of H with f (x) = 2. Assume to the contrary that such a labeling f exists. The assumption implies that f (t) / {1, 2, 3} for t {z1 , z2 , z3 , z4 , y1 , y2 , y3 , y4 }. Also we have, f (zi ) = f (yj ) for i, j = 1, 2, 3, 4 and, for t {y, z }, | f (t1 ) f (t2 ) | 2 and f (t3 ) = f (t4 ) . This will force either {f (z3 ), f (z4 )} = {0, 5} or {f (y3 ), f (y4 )} = {0, 5}. But these two cases are symmetric. So, without loss of generality, we can assume {f (z3 ), f (z4 )} = {0, 5}. Again, by symmetry, we can assume f (z3 ) = 0 and f (z4 ) = 5. This will force + + + + f (v ) = 3. Then, f (t) / {2, 3, 4} for t {v1 , v2 , v4 , v5 , v1 , v2 , v4 , v5 }. + Similarly as before, we have f (vi ) = f (vj ) for i, j = 1, 2, 4, 5 and, for t + {v , v }, | f (t4 ) f (t5 ) | 2 and f (t1 ) = f (t2 ). + Moreover, f (vi ) = f (z3 ) = 0 and f (vi ) = f (z4 ) = 5 for i = 1, 2, 4, 5. + + This forces {f (v1 ), f (v2 )} = {5, 1}. + Then no label is available for v3 , a contradiction. Hence, we do not have a 6-L(2, 1)-labeling f of H such that f (x) = 2.

432

433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450

L(2, 1)-labelings of some families of oriented planar graphs

15

451 452 453 454 455 456 457 458 459 460 461 462

Let G be a graph obtained by gluing a copy of the induced subgraph H [x, y1 , y2 , z1 , z2 ] on each vertex of the directed 5-cycle C5 by identifying each vertex of C5 with the vertex x of H [x, y1 , y2 , z1 , z2 ]. Clearly, G is a cactus. Now, if we 2-dipath 6-L(2, 1)-label G, we need to use at least 5 labels for the C5 inside it. If 2 is not among those 5 labels, then at least one of {4, 5} is among those 5 labels. Now, the H [x, y1 , y2 , z1 , z2 ] glued with the vertex that got label 4 (or 5) clearly must use label 2. Hence, for any 2-dipath 6-L(2, 1)-labeling of the cactus G, we need to use 2 as one of the labels. Now, we construct a new graph C by gluing a copy of H on each vertex of G by identifying that vertex of G with the vertex x of H . Note that C is a cactus that cannot be 2-dipath 6-L(2, 1)-labelled. This completes the proof. Let B be the oriented graph depicted in Figure 1. Then we have: Lemma 14. Let O be an oriented cycle. Given any x V (O) and y V (B ), there exists a homomorphism h : O B such that h(x) = y . Proof. We know that B is 4-nice. Hence it is enough to show that for any oriented 3-cycle T and given any x V (T ) and y V (B ), there exists a homomorphism h : O B such that h(x) = y . In other words, we need to show that for each y V (B ), the 3-cycles in Figure 7 are subgraphs of B , which can easily be checked. Lemma 15. Every oriented catus C admits a homomorphism to B . Proof. Let G be a minimal counterexample to Lemma 15. If there is a degree one vertex v in G such that v N + (u) (or v N (u)) for some u V (G), then G[V (G) \ {v }] is also a cactus. As G is a minimal counterexample, there is a homomorphism f from G[V (G \ {v }] to B . Now, since all the vertices of B have at least one in-neighbor and one out-neighbor, we can extend the homomorphism f to a homomorphism of G to B by mapping v to any vertex x N + (f (u)) (or x N (f (u))). This is a contradiction. Hence there cannot be a degree one vertex in G. No vertex of degree one in G implies at least one cycle C G such that exactly one vertex z of the cycle C has degree greater than 2 (since, by Lemma 14, G cannot be a cycle). Now, G[V (G) \ {V (C ) \ {z }}] is a cactus and, since G is a minimal counterexample, there is a homomorphism f from G[V (G) \ {V (C ) \ {z }}] to B . By Lemma 14, we can extend f to a homomorphism of G to B , a contradiction. This completes the proof. We are now able to prove Theorem 5.

463

464 465

466 467 468 469 470

471

472 473 474 475 476 477 478 479 480 481 482 483 484 485 486

487 488

16 y y y y

S. Sen

Figure 7. Four dierent oriented 3-cycles with respect to the vertex y .

489 490 491

Proof of Theorem 5 The proof follows from Lemma 13, Lemma 15, Lemma 6 and the fact that o 2,1 (B ) = 7 (from Fig: 1).

492

7.

Conclusion

493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508

In this paper we studied 2-dipath and oriented L(2, 1)-span of some planar families of graphs. For the family P of planar graphs we have 17 (P ) 80 where the lower bound is due to Marshall [8] and the upper bound is due to Raspaud and Sopena [10]. In this paper we proved 22 2,1 (P ) o (P ) 83. We proved the upper bound using Raspaud and Sopenas result [10]. But for the lower bound, our proof is independent from Marshalls one [8]. Indeed, using Marshalls result one can only prove 17 o (P ). For the family O of outerplanar graphs we have 9 2,1 (O) o (O) 10. Now, according to this paper, improvements of the form 2,1 (O) = 10 (or o (O) = 10) in the above result will imply 24 2,1 (P ) (or 24 o (P )). We know that there exists an oriented graph on | (P ) | vertices to which every oriented planar graph admits a homomorphism [13]. This paper tells us that 22 labels are not enough to 2-dipath L(2, 1)-label such a graph. We hope one might be able to use this (and other things, especially the congurations used in Marshalls paper [8]) to improve the lower bound for the oriented chromatic number (P ) of the family of planar graphs. References [1] Aardal, K.I. and van Hoesel, S.P.M. and Koster, A.M.C.A. and Mannino, C. and Sassano, A., Models and solution techniques for frequency assignment problems , Annals of Operations Research 153 (2007) 79129. [2] Calamoneri, T. and Sinaimeri, B., L(2, 1)-Labeling of Oriented Planar Graphs (Extended Abstract) , Proc. 12th Italian Conference on Theoretical Computer Science (2010). [3] Chang, G.J. and Chen, J.J. and Kuo, D. and Liaw, S.C., Distance-two labelings of digraphs , Discrete Applied Mathematics 155 (2007) 10071013.

509

510 511 512

513 514 515

516 517

L(2, 1)-labelings of some families of oriented planar graphs

17

518 519

[4] Georges, J.P. and Mauro, D.W., Generalized vertex labelings with a condition at distance two , Congressus Numerantium (1995) 141160. [5] Gon calves, D. and Shalu, MA and Raspaud, A., On Oriented Labelling Parameters , Formal Models, Languages and Applications 66 (2006) 3445. [6] Griggs, J.R. and Yeh, R.K., Labelling graphs with a condition at distance 2 , SIAM Journal on Discrete Mathematics 5 (1992) 586595. [7] Hell, P. and Kostochka, AV and Raspaud, A. and Sopena, E., On nice graphs , Discrete Mathematics 234 (2001) 3951. [8] Marshall, T. H., Homomorphism bounds for oriented planar graphs , J. Graph Theory 55 (2007) 175190. [9] Alexandre Pinlou, An oriented coloring of planar graphs with girth at least ve , Discrete Mathematics 309 (2009) 21082118. [10] Andr e Raspaud and Eric Sopena, Good and Semi-Strong Colorings of Oriented Planar Graphs , Inf. Process. Lett. 51 (1994) 171174. [11] Eric Sopena, The chromatic number of oriented graphs , Journal of Graph Theory 25 (1997) 191205. [12] Eric Sopena, Oriented graph coloring , Discrete Mathematics 229 (13) (2001) 359369. [13] Eric Sopena, There exist oriented planar graphs with oriented chromatic number at least sixteen , Inf. Process. Lett. 81 (2002) 309312. [14] Jan van Leeuwen, Handbook of Theoretical Computer Science, Volume A: Algorithms and Complexity (Elsevier and MIT Press, 1990).

520 521

522 523

524 525

526 527

528 529

530 531

532 533

534 535

536 537

538 539

540