• Tidak ada hasil yang ditemukan

Powers of Graphs and Line Graphs

Dalam dokumen BUKU GRAPHS & DIGRAPH PDF (Halaman 150-162)

A number of graph operations have been defined and studied that have led to several results dealing with Hamiltonian and Eulerian properties. One of the simplest operations is that of the subdivision graph of a graph. The subdi- vision graph S(G) of a graphGis that graph obtained fromGby replacing each edgee=uvofGby a new vertexweand the two new edgesuweandvwe. The subdivision graph ofK4 is shown in Figure 3.15.

IfGis a graph of ordernand size m, then the order ofS(G) isn+mand its size is 2m. Furthermore, S(G) is a bipartite graph with partite sets V(G) andV(S(G))−V(G).

138 CHAPTER 3. EULERIAN AND HAMILTONIAN GRAPHS

.. .. .. . .. . . . .. . .. .. .

......... .............................

.. .. .. . .. . . . .. . .. .. . .. ......

.. .. .. .. . . . . .. . .. .. . .. .......

. .. .. .. . . . . . . .. . .. .. . .. ..... ..

.. .. .. . . . . .. . .. .. . .. ......

. .. . . .. . .. .. . . .. .. . .. ........

.. .. .. . . .. . .. . .. .. .. .........

s s

s s

s s

. ........................................................................................................................................................

.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . . .. .. .. .. . . .. . .. .. .. . .. . .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . . .. . . .. . .. . .. . .. .. .. .. .. .. .. .. ..

................... ............................... ..................................................................................................................................................................

.. .. .. .. .. .. .. .. .. .. . . .. . . .. .. .. .. . . .. . . .. . .. . .. . . .. . . .. . .. .. .. . .. . .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..........

......... .......................................

K4: S(K4) :

Figure 3.15: The subdivision graph of a graph

The Square and Cube of a Graph

Associated with each connected graph of ordernand diameterdis a class of graphs defined in terms of distance. For each positive integer k, the kth powerGk of a graphGis that graph withV(Gk) =V(G) anduv∈E(Gk) if and only if 1≤dG(u, v)≤k. ThusG1 =GandGk =Kn ifk≥d. The graph G2 is also called thesquare ofGwhile G3 is called the cubeof G. A graph with its square and cube are shown in Figure 3.16.

. . .. .. .. .. .. .. .. . . .. .. . .. .. .. . .. .. .. .. . .. .. .. . .. .. .. . .. .. . .. .. . .. .. . .. .. . .. .. . .........................................................

................................................................................................... ..

.. . .. .. . .. .. . .. .. . .. .. .. . .. .. .. . .. .. .. .. .. .. .. . . .. . . .. .. .. .. .. . .. .. . .. .. . .. . .. . .. . .. . .. . . .. . .. . . . .. . . .. . . . . .. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . .. . . . . . .

.................................................................................................. ..

. .. .. . .. .. .. . .. .. .. . .. .. .. .. . .. .. .. .. .. .. .. .. .. .. .. .. .. . .. .. .. . .. .. . .. .. . .. .. . . .. . .. . .. . .. . . .. . . .. . . .. . . . .. . . . .. . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . .

............................................................................................................................................................................................................................................................................. .

. . .. . . . .. . . . . .. . . .. . . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . .. . . .. . .. . .

....................................................................................................................................................................................... .

.. .. . .. .. . .. .. .............................................................................................

.... ....

.............

. .. . .. .. . .. .. . .. . .. .. . .. . .. . .. . .. . .. . . .. . .. . .. . . .. . . .. . .. . . . .. . . .. . . . .. . . . .. . . . .. . . . . . .. . . . . . .. . . . . . . . . .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . .. . . . . . . . . .. . . . . . .. .

.. . .. .. . .. .. . .. .. . .. .. . .. .. . .. .. . .. .. . .. .. . .................................................................................................................................

........

.. .. .. .. .. .. .. .. .. .. .. .. .. . .. .. .. .. . .. .. .. . .. .. . .. .. .. . .. . .. .. .. . .. .. . .. .. . .. .. .................................................... ..

.. . .. . .. .. . . .. .. .. . ........

. .. .. .. . . . . . . .. . .. . .. .. ....

. .. . . .. . .. .. . . .. .. .. . ........ .. .. .. . .. . . . .. . .. . .. .. ..

.... ............................ . .. .. .. . . . . . . .. . .. . .. .......

. .. . .. . . . . . . . . .. .. .. . .......

.. . . .. .. . . . . . .. .. .. .. ........

.. .. . . .. . . . . .. . .. .. . .. ......

. .. .. .. . . . . . . .. .. .. .......... .. .. .. . .. . . .. . . .. . .. ..

..... ........................... .. . .. . .. . . . . . .. .. .. .. .. ....

.. . .. . .. . . . . . .. .. .. .. .. ......

.. .. . . .. . . . . .. . .. .. . .. ....

. .. . .. . . . . . . . . .. .. .. . .. ....

.. .. .. . .. . . .. . . .. . .. .. .....

. .. .. .. . . . . . . .. .. .. .. ..

...... ............................. .

.. . . .. . .. . .. . .. .. ..

......... ........................... .... ...........

..... .......

......

....................................................

......................................................................................................................................................................

.. ................................................................................................................

.........

...... ...........................................................................................................

....... ......

. .. .............................................................................................................

.. .. .. .. .. . .. . .. . .. .. .. .. .. .. .. .. . .. .. .. .. .. .. .. .. .. .. . .. .. .. .. .. .. .. .. .. .. . .. .. .. .. .. .. .. .. .. .. . .. . .. . .. .. .. .. . ............ ................................................................................................

...

.........

G3: G2:

G:

Figure 3.16: A graph and its square and cube

If G is a nontrivial connected graph of order n and k is an integer with 1≤k≤n−1, thenGk isk-connected (see Exercise 4). Since thekth powerGk (k≥2) of a connected graphGcontainsGas a subgraph (as a proper subgraph if G is not complete), it follows that Gk is Hamiltonian if G is Hamiltonian.

WhetherGis Hamiltonian or not, for a connected graphGof order at least 3 and for a sufficiently large integerk, the graphGk is Hamiltonian sinceGd is complete ifGhas diameterd. It is therefore natural to ask for the minimumk for whichGk is Hamiltonian. Certainly, for connected graphs in general,k= 2 will not suffice since ifG=S(K1,3) is the graph of Figure 3.16, thenG2 is not Hamiltonian. The graphG3isHamiltonian, however.

In fact, the cube of every connected graph of order at least 3 is Hamiltonian.

Indeed, a stronger result exists, discovered by Milan Sekanina [202] and later, but independently, by Jerome Karaganis [129].

Theorem 3.26 IfGis a connected graph, thenG3is Hamiltonian-connected.

Proof. IfH is a spanning tree ofGandH3 is Hamiltonian-connected, then G3is Hamiltonian-connected. Hence it suffices to prove that the cube of every tree is Hamiltonian-connected. To show this, we proceed by induction on n, the order of the tree. For small values ofnthe result is obvious.

Assume for all treesHof order less thannthatH3is Hamiltonian-connected and letT be a tree of ordern. Letuandvbe any two vertices ofT. We consider two cases.

Case1. uandvare adjacent inT. Lete=uv, and consider the forestT−e. This forest has two components, one tree Tu containing u and the other tree Tv containingv. By hypothesis,Tu3andTv3are Hamiltonian-connected. Letu1 be any vertex ofTu adjacent tou, and let v1 be any vertex ofTv adjacent to v. IfTu orTv is trivial, we defineu1=uor v1=v, respectively. Note thatu1 andv1are adjacent inT3sincedT(u1, v1)≤3. LetPube a Hamiltonianu−u1 path (which may be trivial) ofTu3 and letPv be a Hamiltonianv1−v path of Tv3. The path formed by beginning withPufollowed by the edgeu1v1 and then the pathPv is a Hamiltonianu−vpath ofT3.

Case 2. u and v are not adjacent in T. Since T is a tree, there exists a unique path between every two of its vertices. LetP be the uniqueu−vpath of T, and let f = uw be the edge of P incident with u. The graph T −f consists of two trees, one treeTu containinguand the other treeTwcontaining w. By hypothesis, there exists a Hamiltonian w−v path Pw in Tw3. Let u1 be a vertex ofTu adjacent to u, or let u1 =uifTu is trivial, and let Pu be a Hamiltonianu−u1path inTu3. BecausedT(u1, w)≤2, the edgeu1wis present in T3. Hence the path formed by starting withPu followed by u1w and then Pwis a Hamiltonianu−vpath ofT3.

It is, of course, an immediate consequence of Theorem 3.26 that for any connected graph G of order at least 3, its cube G3 is Hamiltonian. Although it is not true that the square of every connected graph of order at least 3 is Hamiltonian, it was conjectured independently by Crispin Nash-Williams and Michael D. Plummer that for 2-connected graphs this is the case. In 1974, Her- bert Fleischner [82] verified this conjecture. More recent proofs of this theorem have been given by by Stanislav ˇRiha [183] in 1991 and Agelos Georgakopoulos [93] in 2009.

Theorem 3.27 If G is a2-connected graph, thenG2 is Hamiltonian.

A variety of results strengthening (but employing) Fleischner’s work have been obtained. For example, Gary Chartrand, Arthur Hobbs, Heinz Jung, S. F.

Kapoor and Crispin Nash-Williams [37] showed that the square of a 2-connected graph is Hamiltonian-connected.

Theorem 3.28 If G is a2-connected graph, thenG2is Hamiltonian-connected.

Proof. SinceG is 2-connected, G has order at least 3. Let uand v be any two vertices ofG. LetG1, G2, . . . , G5be five distinct copies ofGand letuiand

140 CHAPTER 3. EULERIAN AND HAMILTONIAN GRAPHS vi (i= 1,2, . . . ,5) be the vertices inGi corresponding touandvinG. Form a new graphF by adding to the unionG1+G2+· · ·+G5two new verticesw1and w2and ten new edgesw1ui andw2vi(i= 1,2, . . . ,5). The graphF is shown in Figure 3.17. Clearly, neitherw1norw2is a cut-vertex ofF. Furthermore, since each graphGi is 2-connected and contains two vertices adjacent to vertices in V(F)−V(Gi), no vertex ofGi is a cut-vertex ofF. HenceF is 2-connected.

.. .. . .. .. .. . .. . .. .. .. . .. .. . .. .. .. . .. .. .. .. . .. .. .. .. .. .. .. .. .. . .. .. . .. . . .. . . . . . . . . . . . . . .. . . .. . . .. .. . .. .. .. .. .. .. .. .. .. .. . .. .. .. . .. .. .. . .. .. . .. .. .. . .. .. . .. .. . .. .. .. . .. . .. .. . .. ..........................................

...................................................................... .......................................................................................................................................................................................................................

...................................................................... ............................................................................................................................................................................................................................................................................................. .................................................................................................................................................................................................................................................................................................. ............................................................................................................................................................................................................................................................................................

s

s

s s s s

s s s s

s s

...............................................................

.

....................................................................................................................................................................................................

.................................

............

...............

......

................

................................................................................................................................................................................

...............................................................

G1 G2 G3 G4 G5

w1

w2

u1 u3 u5

v1 v3 v4

u4

v2 v5

F :

u2

Figure 3.17: The graphF in the proof of Theorem 3.28

By Theorem 3.27,F2has a Hamiltonian cycleC, which, of course, contains w1 andw2. At least one of the graphsGi, sayG1, contains no vertex adjacent tow1orw2 onC. Sinceu1andv1are the only vertices ofG1 adjacent onCto vertices not inG1, it follows thatChas au1−v1path containing all vertices of G1. ThusG21 has a Hamiltonianu1−v1 path, which implies that G2contains a Hamiltonianu−v path.

The Line Graph of a Graph

The most familiar graph operation of a graph is that of the line graph. The line graphL(G) of a graphGis that graph whose vertices can be put in one-to- one correspondence with the edges ofGin such a way that two vertices ofL(G) are adjacent if and only if the corresponding edges ofGare adjacent. A graph and its line graph are shown in Figure 3.18 where the vertexui (1≤i≤6) of L(G) corresponds to the edgeei ofG.

It is relatively easy to determine the number of vertices and the number of edges in the line graphL(G) of a graphGin terms of easily computed quantities in G. Indeed, if G is a graph of order n and size m with degree sequence d1, d2, . . . , dn andL(G) is a graph of ordern and sizem, thenn =m and

m=

n

X

i=1

di

2

. .. .. .. . . . . . . .. . .. . .. ..

..... .............................

.. .. . . .. . . . . .. . .. .. .. . .......

.. . .. . .. . . . . . .. .. .. .. ..

..... .............................

.. .. .. . .. . . .. . .. .. .. .. ......

.. . .. .. . . . . . . .. .. .. .. ........

.. . .. . .. . . . . . .. .. .. .. ....... .

.. .. .. . . . . . . .. .. .. .. .. ......

.. . .. . .. . . . . . .. .. .. .. .....

. .. .. .. . . . . . . . .. .. .. . .....

..

..................... .....................................................................

..... ....................

......................................................................

............................... .......................................................................... ....................................................................................... .........

..... .....

......... ...

......... ..............

............... .. . .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. . .. . . .. .. .. .. . .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. . .. .. .. .. .

u6 u3 e3

e1 e2

e5 G:

e6

L(G) : e4

u1

u4

u2

u5

Figure 3.18: A graph and its line graph

since each edge of L(G) corresponds to a pair of adjacent edges of G. For each nontrivial connected graph G, its line graph L(G) is also connected (see Exercise 10). For the setGof all connected graphs and the setGof all nonempty connected graphs, we may think of L as a function then, namelyL :G → G. Hassler Whitney [238] showed that this function is very nearly injective.

Theorem 3.29 Let G1 and G2 be nontrivial connected graphs. If L(G1) ∼

= L(G2), then G1

=G2 unless one ofG1 andG2 isK3 and the other isK1,3. The function L : G → G is not close to being surjective, however. The graphK1,3is one of many graph that is not isomorphic to the line graph of any graph. Suppose that there is a graphH such that L(H) =K1,3. Then H is a graph of size 4 containing an edge that is adjacent to the other three edges, no two of which are adjacent to each other. Such a graph H does not exist and soK1,3is not the line graph of any graph. Indeed, Lowell W. Beineke [10]

obtained the following result.

Theorem 3.30 A graphGis isomorphic to the line graph of some graph if and only if none of the graphs of Figure 3.19is isomorphic to an induced subgraph of G.

We turn to the problem of determining characteristics possessed by a graph that yield certain Hamiltonian properties of its line graph. Frank Harary and Crispin Nash-Williams [112] characterized those graphs having a Hamiltonian line graph. A circuit C in a graphGis called adominating circuitif every edge ofG either belongs toC or is adjacent to an edge of C. Equivalently, a circuitCin a graphGis a dominating circuit if every edge ofGis incident with a vertex ofC.

Theorem 3.31 Let G be a graph without isolated vertices. Then L(G) is Hamiltonian if and only ifG=K1,ℓfor someℓ≥3orGcontains a dominating circuit.

Proof. IfG=K1,ℓ for someℓ≥3, then L(G) is Hamiltonian sinceL(G) = K. Suppose, then, thatGcontains a dominating circuit

C= (v1, v2, . . . , vt, v1).

142 CHAPTER 3. EULERIAN AND HAMILTONIAN GRAPHS

.. . .. . .. . . . . . .. .. .. .. .

....... .............................

.. . .. . .. . . . . . .. .. .. .. . .......

. .. .. .. . . . . . . . .. .. .. . ..

...... ............................

. .. .. .. . . . . . . .. .. .. .. .. ......

. .. .. . . . . . . . . . .. .. ..

...... ............................. .

.. . .. . .. . . .. . .. .. .. .. . ......

.. . .. . .. . . . . . .. .. .. .. .

....... ............................. .............................

. .. .. .. . . . . . . . .. .. . .. ........

.. .. .. . .. . . .. . .. . .. . .. .......

. .. .. . . . . . . . . . .. .. .. .. . ......

. .. .. .. . . . . . . . .. .. .. . .. ...... . .. .. . . . . . . . . . .. .. ..

........ .............................

.. .. .. . .. . . .. . .. .. .. .. .......

.. .. .. . .. . . . .. . .. .. . .. ....... .. . .. . .. . . . . . .. .. .. .. .

....... ............................. ..

.. .. .. . . . . . .. .. .. .. . ...... .

.. .. .. . . . . . . .. .. .. ..........

. .. .. . . . . . . . . . .. .. .. .. . ......

. .. .. .. . . . . . . .. . .. . .. .. ....

.. . .. .. . . . . . . .. .. .. .. .. ....

. .. .. .. . . . . . . .. . .. . .. .. ......

.. . .. .. . . . . . . .. .. .. .. .. ......

. .. . .. . . .. .. . . .. .. .. . ........

. .. .. .. . . . . . . .. .. .. .. ..

...... .............................

. .. .. . . . . . . . . . .. .. .. .

........ ............................

.. .. . . .. . . . . .. . .. .. .. .

....... ..............................

. .. .. .. . . . . . . .. . .. . .. ......

.. . .. .. . . . . . . .. .. .. .. ....... ..

.. .. . . .. . .. . .. .. .. .. ....

.. .. .. . . .. .. . . .. .. .. . .. ...... ..

.. .. .. . . . . .. . .. .. . ........

.. .. .. . .. . . .. . .. .. .. .. . ......

. .. .. .. . . . . . . . .. .. .. .

........ .............................

.. .. .. . . .. .. . . .. .. .. . .. ......

. .. . .. . .. . . .. . .. .. .. ..

...... .............................

.. .. .. . .. . . . .. . .. .. . .. .. .....

.. .. .. . .. . . . .. . .. .. . .. ....... ..

.. .. . .. . . . .. . .. . .. .. ....... ..........

........ ................................................... ....................

............................................................ .

........................................................... ................

.....

...................

....................................

.. ...................................................................................................................

................ .........

................................................................................. .............................

........................................................................................................................

............................................................................. ..........................

.................. ...........

. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. .. . ..

............................................................................... ......... .............

...

.............. ..........

. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. .. . ..

............... ......... ................................................................................. ............... ..............................

......... ................

.... ......................

. .. .. .. . .. .. .. . .. . .. .. . .. .. .. . .. .. .. . .. . .. .. .. . .. .. .. . .. .. . .. .. . .. .. .. . .. .. . .. .. . .. .

...................................................... ... .............

...

. . ....................................... ............................................................

...

...

......................................................................................................

................................................................

. .....................................................................

. .. .. .. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. . .. .. .. .. . .. ..

. .

.

.............................................................................. ...... ................

...... ......

................................................................. .. .. . . .. . .. . .. . .. .. .. .. .. .. .. .. .. .. .. .. . . .. . .. . .. . .. .. .. .. .. .. ..

Figure 3.19: The induced subgraphs not contained in any line graph

It suffices to show that there exists an orderings:e1, e2, . . . , emof themedges ofGsuch thatei andei+1are adjacent edges ofG, for 1≤i≤m−1, as aree1 andem, since such an orderingscorresponds to a Hamiltonian cycle ofL(G).

Begin the orderingsby selecting, in any order, all edges ofGincident withv1 that are not edges of C, followed by the edgev1v2. At each successive vertex vi, 2≤i≤t−1, select, in any order, all edges of G incident with vi that are neither edges of C nor previously selected edges, followed by the edge vivi+1. This process terminates with the edgevt−1vt. The orderingsis completed by adding the edge vtv1. Since C is a dominating circuit of G, every edge of G appears exactly once in s. Furthermore, consecutive edges of sas well as the first and last edges ofsare adjacent inG.

Conversely, suppose thatGis not a star butL(G) is Hamiltonian. We show that G contains a dominating circuit. Since L(G) is Hamiltonian, there is an orderings:e1, e2, . . . , emof themedges of G such thateiandei+1are adjacent edges ofGfor 1≤i≤m−1, as aree1andem. For 1≤i≤m−1, letvi be the vertex ofGincident with bothei andei+1. (Note that 1≤k6=q≤m−1 does not necessarily imply that vk 6=vq.) Since Gis not a star, there is a smallest integerj1exceeding 1 such thatvj1 6=v1. The vertexvj11is incident withej1, the vertex vj1 is incident with ej1 and vj11 =v1. Thus, ej1 =v1vj1. Next, let j2 (if it exists) be the smallest integer exceeding j1 such that vj2 6= vj1. The vertex vj21 is incident with ej2, the vertex vj2 is incident with ej2 and vj21 =vj1. Thus, ej2 =vj1vj2. Continuing in this fashion, we finally arrive at a vertex vjt such that ejt = vjt1vjt, wherevjt = vm−1. Since every edge of Gappears exactly once ins and since 1< j1< j2 <· · ·< jt≤m−1, this

construction yields a trail

T= (v1, ej1, vj1, ej2, vj2, . . . , vjt1, ejt, vjt =vm−1)

inGwith the properties that (i) every edge ofGis incident with a vertex ofT and (ii) neithere1noremis an edge ofT.

Letwbe the vertex of Gincident with both e1 andem. We consider four possible cases.

Case1. w=v1=vm−1. ThenT itself is a dominating circuit ofG.

Case2. w=v1 andw6=vm−1. Sinceemis incident with bothwandvm−1, it follows thatem =vm−1w =vm−1v1. ThusC = (T , em, v1) is a dominating circuit ofG.

Case 3. w= vm−1 andw 6=v1. Since e1 is incident with both w and v1, we have thate1=wv1 =vm−1v1. ThusC= (T , e1, v1) is a dominating circuit ofG.

Case4. w6=vm−1 andw6=v1. Sinceemis incident with bothwandvm−1, it follows thatem=vm−1w. Sincee1is incident with both wandv1, we have that e1 =wv1. Thus v1 6= vm−1, and C = (T , em, w, e1, v1) is a dominating circuit ofG.

As a consequence of Theorem 3.31, ifGis either Eulerian or Hamiltonian, then L(G) is Hamiltonian. In fact, successively taking the line graph of a connected graph has some interesting consequences.

For a nonempty graphG, we writeL0(G) to denoteGandL1(G) to denote L(G). For an integer k ≥ 2, the iterated line graph Lk(G) is defined as L(Lk−1(G)), whereLk−1(G) is assumed to be nonempty. The following result is due to Gary Chartrand and Curtiss E. Wall [41].

Theorem 3.32 IfGis a connected graph such thatdegv≥3 for every vertex v of G, thenL2(G)is Hamiltonian.

Proof. Let v be a vertex of G, where degv = r ≥ 3. Then in L(G), the edges incident withv give rise to a subgraphHv, whereHv

=Kr. LetCv be a Hamiltonian cycle inHv. LetH be the spanning subgraph ofL(G) defined by

V(H) =V(L(G)) andE(H) =S

v∈V(G)E(Cv).

Then H is connected and every vertex of H is even. Consequently,H is Eu- lerian and so H is a dominating circuit of L(G). By Theorem 3.31, L2(G) is Hamiltonian.

If G = Pn, where n ≥ 2, then L(G) = Pn−1. Thus Ln−1(G) = P1 and Lk(G) is not defined fork ≥n. If G=K1,3, thenL(G) =C3. If G=Cn for somen≥3, then L(G) =Cn andLk(Cn) =Cn for every nonnegative integer

144 CHAPTER 3. EULERIAN AND HAMILTONIAN GRAPHS k. Suppose then that G is a connected graph that is not a path, K1,3 or a cycle. Then ∆(G)≥3. Suppose thatGcontains a vertexuof degree 1. Then Gcontains au−v pathP = (u=u0, u1, . . . , u=v) of minimum lengthℓ≥1 where degv ≥ 3. This gives rise to an x0−x path P of length ℓ in L(G), where x0 corresponds to u0u1, xcorresponds to uℓ−1u, degx0 = 1 if ℓ ≥ 2, and degx≥3. By successively taking the line graph of G, L(G), L2(G) and so on, we eventually arrive at a graph Lk(G) for some k where there are no end-vertices. If a connected graphH contains no end-vertices but does contain vertices of degree 2, then G contain a y −z path (or y −z cycle if y = z) Q= (y=y0, y1, . . . , yt=z) wheret≥2, degyi= 2 for 1≤i≤t−1, degy≥3 and degz≥3. This gives rise to aw−xpathQ= (w=w1, w2, . . . , wt=x) of lengtht−1 inL(G) wherew1corresponds to the edgey0y1inG,wtcorresponds to the edgeyt−1yt, degwi= 2 for 2≤i≤t−1, degw1≥3 and degwt≥3. By successively taking the line graph ofH,L(H), L2(H) and so on, we arrive at a graphLr(H) in which all vertices have degree at least 3. This is illustrated for the graphGof Figure 3.20. It then follows that there is a sufficiently large integer s such that the degree of every vertex ofLs(G) is at least 3. By the discussion above and Theorem 3.32, we then have the following.

. .. .. .. . . . . . . . .. .. .. .

....... ............................. ............................. ............................. .............................

. .. .. .. . . . . . . . .. .. .. . .......

. .. . .. . .. . . .. . .. .. ..

........ .............................

. .. .. .. . . . . . . . .. .. .. . ..

...... ............................. ............................ ............................. .

.. . .. . . . . . . . . .. .. .. .. ....... .

.. .. .. . . . . . . . .. .. .. . .. .....

.. . .. . .. . . . . . .. .. .. .. .. .....

.. .. .. .. . . . . .. . .. .. . .. .. .....

.. .. .. . .. . . .. . .. .. .. .. ......

.. .. .. .. . . . . . .. .. .. . ......... .

.. .. .. . . . . . . . .. .. . .. ....... ..

.. .. . . .. . .. . .. .. .. .. .....

.. .. .. . .. . . . .. . .. . .. .......

.. . .. .. . . . . . . .. .. .. .. ..

...... ...........................

.. .. .. . . .. . .. . .. .. ..

....... ............................. ........................... ............................. .

.. .. .. . . . . . . . .. .. .. . ........

. .. . . .. . .. . .. . .. .. .. ..

....... ............................. ........................... ........................... ..

. .. . .. . . . . . .. .. .. .. .. ..... ..

. .. . .. . . . . . .. .. .. .. . .......

.. .. .. . .. . . .. . .. . .. . .........

. ................................................. . .. .. .. .. .. .. .. .. .. .. .. .. .. .. . . .. .. . . .. . . .. . .. .. . .............

....................................................

....................................................... ...................... .................................

.. .. .. . .. .. .. .. .. .. .. .. . .. .. .. .. .. . .. . . .. .. .. .. .. . .. .. .. .. .. .

.. .................................................... ..........................

...........................

. .. . .. .. .. .. . .. .. .. .. .. . .. .. .. .. . .. .. .. .. .. . .. .. .. .. . .. .. ..

...............................

............

........... . ..................................

..................................

. ........................................................................... .....

..........

............ ..................................... .

..............................................................................................................

................

.. .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. . .. .. .. .. . .. .. .. . .. .. .. . .. .. .. . .. .. .. .

........................................................................................ ...................................................

................................................................... ..............................

G:

L2(G) : L3(G) :

L(G) :

Figure 3.20: A graph and iterated line graphs

Theorem 3.33 If Gis a connected graph that is not a path, then there exists an integern0 such thatLk(G)is Hamiltonian for every integerk≥n0.

The Total Graph of a Graph

A graph operation related to the line graph is the total graph. Thetotal graph T(G) of a graph G is that graph whose vertices can be put in one- to-one correspondence with the elements of the set V(G)∪E(G) such that two vertices of T(G) are adjacent if the corresponding elements in G are two adjacent vertices, two adjacent edges or an incident vertex and edge. A graph H and its total graphT(H) are shown in Figure 3.21.

. .. . . .. .. .. .. .. . . .. .. . .. .. .. .. .. . .. .. .. . .. .. . .. .. .. .. . .. . .. .. .. . .. .. . ..................................................................

.................................................................................................... .

.. . .. .. .. . .. . .. .. .. . .. .. . .. .. .. . .. .. .. .. .. .. .. .. . . .. . . .. .. .. .. . .. .. .. . .. .. . .. . .. . .. . .. . .. . . .. . . .. . . . .. . . . .. . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. .. .. . . .. . . . . .. . .. .. .. . ......

.. .. .. . .. . . . .. . .. . .. .. .......

.. .. . . .. . . . . .. . .. .. .. . .....

. .. .. .. . . . . . . . .. .. .. .

........ ............................. .. .. .. . .. . . .. . .. .. .. .. .. .....

.. . .. . .. . . . . . .. .. .. . ......... ..

.. .. . . .. .. . . .. .. .. ......... .. .. .. .. . . . . .. . .. .. . ..

..... ...........................

.. . .. .. . . . . . . .. .. .. .. .. ......

........ .......

.................

....................................................

............................................................................................................................................................................ ................................................................................................................ ..

........................................................................... . .. . . .. . . .. . . .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..

. T(H) :

H :

Figure 3.21: A graph and its total graph

One might observe thatT(H)∼=G2for the graphH of Figure 3.21 and the graphGof Figure 3.16. We saw thatG2is not Hamiltonian. Thus, for the graph H of Figure 3.21,T(H) is not Hamiltonian. This, in fact, is not surprising for if F is any graph, thenT(F)∼

= [S(F)]2, that is, the total graph ofF is the square of the subdivision graph ofF. Since the graphG of Figure 3.16 is isomorphic toS(K1,3) and the graphH of Figure 3.21 is isomorphic toK1,3, it follows that T(H)∼

=G2, whereGis the graph of Figure 3.16.

While the total graph of a nontrivial connected graphGneed not be Hamil- tonian, the same cannot be said forT(T(G)).

Theorem 3.34 If Gis a nontrivial connected graph, then T(T(G))is Hamil- tonian.

Proof. First, we considerT(G). SinceS(G) is a nontrivial connected graph, [S(G)]2 is 2-connected and T(G) ∼= [S(G)]2, it follows that H = T(G) is 2- connected. HenceS(H) is 2-connected. By Theorem 3.27, [S(H)]2 is Hamilto- nian. ThusT(H)∼

= [S(H)]2and so T(T(G)) is Hamiltonian.

Exercises for Section 3.3

1. Determine all those connected graphsGfor whichS(G) is Eulerian.

2. Determine all those connected graphsGfor whichS(G) is Hamiltonian.

3. Show that ifGis a connected graph of ordernand sizem, thenα(S(G)) = munless Gbelongs to a familiar class of graphs.

146 CHAPTER 3. EULERIAN AND HAMILTONIAN GRAPHS 4. Show that if G is a connected graph of ordern≥2 and k is an integer

with 1≤k≤n−1, thenGk isk-connected.

5. Show that if G is a connected graph of diameterℓ and 1≤k ≤ℓ, then diam(Gk) =⌈ℓ/k⌉.

6. A graphH is called asquare rootof a connected graphGifH2=G. (a) Give an example of a connected graph with two non-isomorphic

square roots.

(b) Give an example of a connected graph with a unique square root.

7. Show that the graphG2 of Figure 3.16 is not Hamiltonian.

8. Prove that if v is any vertex of a connected graphGof order at least 4, thenG3−v is Hamiltonian.

9. Prove that ifGis a self-complementary graph of order at least 5, thenG2 is Hamiltonian-connected.

10. Prove that the line graph of every nontrivial connected graph is connected.

11. Determine a formula for the number of triangles in the line graph L(G) in terms of quantities inG.

12. Prove thatL(G) is Eulerian ifGis Eulerian.

13. (a) Find a necessary and sufficient condition for a graphGto have the property that G∼=L(G).

(b) Find a necessary and sufficient condition for a graphGto have the property that L(G)∼=L2(G).

14. For each of the following, prove or disprove.

(a) IfGis Hamiltonian, thenG2 is Hamiltonian-connected.

(b) IfGis connected andL(G) is Eulerian, thenGis Eulerian.

(c) IfGis Hamiltonian, thenL(G) is Hamiltonian-connected.

(d) IfGhas a dominating circuit, thenL(G) has a dominating circuit.

15. Prove that ifG is a connected graph andL3(G) is Eulerian, thenL2(G) is Eulerian.

16. Give an example of a connected graph G such that degv ≥3 for every vertexvof Gbut L(G) is not Hamiltonian.

17. (a) Show that if G is a k-edge-connected graph, k ≥ 2, then L(G) is k-connected.

(b) Show that if G is a k-edge-connected graph, k ≥ 2, then L(G) is (2k−2)-edge-connected.

18. Show that there exists a graphGthat is not isomorphic to the total graph of any graph.

19. (a) Prove that ifGis a nontrivial connected graph, thenT(G2) is Hamil- tonian.

(b) Prove that if G is a nontrivial connected graph, then (T(G))2 is Hamiltonian.

Chapter 4

Digraphs

While there are many concepts in digraphs that are analogues to concepts we have encountered in graphs, there are also concepts that are quite unique to digraphs. We consider many of these in this current chapter, beginning with the most studied type of connectedness for digraphs.

Dalam dokumen BUKU GRAPHS & DIGRAPH PDF (Halaman 150-162)