• Tidak ada hasil yang ditemukan

Directory UMM :Journals:Journal_of_mathematics:DM:

N/A
N/A
Protected

Academic year: 2017

Membagikan "Directory UMM :Journals:Journal_of_mathematics:DM:"

Copied!
13
0
0

Teks penuh

(1)

Hamiltonian Virus-Free Digraphs

Digrafos Libres de Virus Hamiltonianos

OsarOrdazandLeidaGonzalez(*)

(oordazanubis.iens.uv.ve)

DepartamentodeMatematiaand

CentrodeIngenieradeSoftwareYSistemasISYS.

FaultaddeCienias. UniversidadCentral deVenezuela

Ap. 47567,Caraas1041-A,Venezuela.

IsabelMarquez

UniversidadCentroidentalLisandroAlvarado

DeanatodeCienias. DepartamentodeMatematia.

Barquisimeto,Venezuela.

DomingoQuiroz(dquirozusb.ve)

UniversidadSimonBolivar. DepartamentodeMatematia.

Ap. 89000,Caraas1080-A,Venezuela.

Abstrat

A hamiltonian virus is a loal onguration that, if present in a

digraph, forbids thisdigraph to have a hamiltonianiruit.

Unfortu-nately,therearenon-hamiltoniandigraphsthatarehamiltonian

virus-free. Somefamiliesofthesedigraphswillbedesribedhere. Moreover,

problemsandonjeturesrelatedtohamiltonianvirus-freedigraphsare

given.

Keywordsand phrases: digraph,hamiltoniandigraph,hamiltonian

virus.

Resumen

Un virus hamiltoniano es una estrutura loal que, estando

pre-senteen un digrafo, impideque este tenga un iruito hamiltoniano.

Desafortunadamente, existen digrafos no hamiltonianos sin virus

ha-miltonianos. Algunasfamiliasdeestosdigrafossondesritasaqu. Mas

aun, seplanteanproblemasy onjeturas relativas adigrafos sin virus

hamiltonianos.

Palabras y frases lave: digrafos,digrafoshamiltonianos, virus

ha-miltonianos.

Reeived: 1999/06/04.Revised: 1999/10/28. Aepted: 1999/11/15.

(2)

1 Introdution and terminology

Theimportane ofhamiltonianviruses[1,8℄istheirrelationwiththe

\erti-ation" ofnon-hamiltonian digraph families, i.e. they arenon-hamiltonian

ifandonlyiftheyhaveahamiltonianvirus. Forexample,balanedbipartite

digraphsarehamiltonianifand onlyiftheyarehamiltonianvirus-free. This

paperhasmultiple goals:

To identify non-hamiltonian digraph families whih are hamiltonian

virus-free,anddigraphfamiliesthat arehamiltonianifandonlyifthey

arehamiltonianvirus-free.

To haraterizedigraphs that donot ontainhamiltonian virusesof a

givenorder.

To present and disuss problems and onjetures related to expeted

propertiesofhamiltonianvirus-freedigraphfamilies.

1.1 Terminology

The terminologydesribedin what followsis taken textuallyfrom [2℄and it

will beusedthroughoutthewholepaper.

Invariants areintegerorbooleanvaluesthat arepreservedunder

isomor-phism. Wewillbeusingthefollowinginvariants,relationsbetweeninvariants,

theorems anddigraphexamples. LetD=(V(D);E(D))beadigraph.

Integer invariants

nodes: numberofnodesofadigraph.

ars: numberofarsofadigraph.

alpha2: maximumsizeofasetofnodeswhihinduesnoiruitoflength2.

alpha0: maximumsizeofasetofnodesinduingnoar.

woodall: min fd +

(x)+d (y) : (x;y) 2= E(D);x 6= yg (if alpha2 = 1, then

woodall =2nodes byonvention.)

minimum: minfmindegpositive,mindegnegativeg.

mindegpositive minfd +

(x):x2V(D)g.

mindegnegative minfd (x):x2V(D)g.

Boolean invariants

hamiltonian: thedigraphontainsahamiltonian iruit.

traeable: thedigraphontainsahamiltonianpath.

(3)

bipartite: its vertex set is partitioned into twosubsets X and Y suh that

eaharhasonevertexin X andanotherin Y.

antisymmetri: itdoesnotontainairuitoflengthtwo.

(1,1)-fator: itontainsaspanningsubdigraphHsuhthatd +

H (x)=d

H (x)=

1forallverties.

Relationsbetween invariants

R

11

: minimum 2 ^ nodes4=)hamiltonian.

R

31

: antisymmetri=)arsnodes(nodes 1)/2.

Theorems and onjetures

Theorem 51: k-onneted^ (alpha0 k) =) (1,1)-fator. Best result: see

D

20 .

Theorem64: antisymmetri^(nodes2h+2)^(h2)^(minimumh)

=)hamiltonian.

Theorem 65: antisymmetri ^ (nodes 6) ^ (woodall nodes 2) =)

hamiltonian.

Theorem66: antisymmetri^(h5)^(minimumh)^(nodes2h+5)

=)hamiltonian.

Theorem67: antisymmetri^2-onneted^[arsnodes(nodes 1)/2 2℄

=)hamiltonian. Best result: seeD

20 :

Theorem 77: (nodes = 2a+1) ^ (minimum a) =) hamiltonian _D

5 _

D

6 _D

7 _D

8 .

Theorem 78: r-diregular^(nodes=2r+1)=)hamiltonian _D

5 _D

6 .

(4)

2 Hamiltonian viruses

A hamiltonian virus is a loal onguration that, if present in a digraph,

forbidsthisdigraphtohaveahamiltonianiruit.

Theorem1 ([1℄). Let H = (V(H);E(H)) be a proper indued subdigraph

of a given digraph D = (V(D);E(D)). A 3-uple (H;T +

;T ), where T +

=

fx 2 V(H) : d +

H

(x) = d +

D

(x)g and T = fx 2 V(H) : d

H

(x) = d

D (x)g,

is a hamiltonian virus if and only if for every set of disjoint direted paths

P

1 ;:::;P

r

overing V(H)there existsapath P

j =x

1

j :::x

q(j)

j

, withq(j)1

suh that either x 1

j

2 T or x q(j)

j 2 T

+

. The order of a hamiltonian virus

(5)

Inwhatfollowsa3-uple(H;T +

;T )ispresentinadigraphDifandonly

if there exists in D a proper indued subdigraph H 1

isomorphi to H (for

onveniene weidentify H 1

is a hamiltonian virusof a given digraphD; then wemust haveT +

agivendigraph, thenthere existsadigraphD where (H;T +

;T )is present

andD ishamiltonian.

Theorem2. If adigraphofordernisfreeof hamiltonianvirusesoforderh

for someh with 2h<n; then it has no hamiltonian viruses of order less

thanh.

Equivalently,fromahamiltonianvirusoforder1hn 2weanbuild

ahamiltonianvirusoforderh+1:

Proof. Let us reason ab absurdo. Let D be a digraph of order n and 2

h n 1: Assuming D ontains ahamiltonian virus (H;T +

;T ) of order

h 1. Letx2V(D)nV(H)andH

1

thesubdigraphinduedbyV(H[fxg)

in D. It islear thatthe 3-uple(H

is presentin D. Now,weshall seethat (H

1

)isahamiltonianvirus.

Let P

1 ;:::;P

r

be a set of disjointdireted paths overingV(H

1

): Without

lossofgeneralityweansupposethat x2V(P

1

):Weonsider threeases:

Case1P

isahamiltonianvirusthereexistsapathP 1

:Thereforex 1

. Thissituationwillbetreatedas

Case1.

. Therefore (H

1

) is a hamiltonian

virus oforder h. Aontradition.

Corollary 1. Ifadigraphofordernhashamiltonianvirusesthenitontains

(6)

FromCorollary1wehave:

(x)gisnotahamiltonianvirus thenD is

hamiltonianvirus-free. Moreover,ifD xisnotahamiltonianvirusforsome

x,thenforeahhamiltonianvirus(H;T +

;T )ofD wehavex2V(H):

DigraphD

6

showsthatthereexistnon-hamiltoniandigraphshamiltonian

virus-free. In D

6

hamiltonian viruses of order 6 are not present. Then by

Corollary1D

6

doesnothaveviruses.

FromTheorem1wehave:

Remark 2. Ahamiltonianvirus-freedigraphD hasthefollowingstruture:

for eahvertexxtheremainingpartD x hasaoveringbyvertexdisjoint

pathsP

1 ;:::;P

r

suhthateahoneof themmakesairuitwithx.

InthenextlemmaD[S℄denotesthesubdigraphinduedbyS.

Lemma1. Let (H;T

;T )isahamiltonian viruspresentin D.

Proof. LetP

1 ;:::;P

r

beasetof disjointdireted pathsoveringV(H S):

Thenforanysetofdisjointdireted pathsP

r+1 ;:::;P

s

overingV(D[S℄)we

havethatP

1

isasetofdisjointdiretedpathsovering

V(H). Sine(H;T +

;T )is ahamiltonianvirus, there exists apath P

j

Theorem3. Let D be an antisymmetridigraph with minimum 2. Then

D isfreeof hamiltonian virusesoforder 4.

Proof. Let us reason ab absurdo. Let (H;T +

j3(similarly,jT j3). Byasimpleinspetionontherelative

positionsofthearsinH,weandeduethateitherthereexistsasymmetri

ar or(H;T +

;T )isnotahamiltonianvirus. Aontradition. Thereforeit

mustbe jT +

j 1 and jT j 1. If T +

= ;then T 6= ;, henethere are

disjoint direted paths overingV(H) that do not verify the onditions for

(H;T +

;T ) to bea hamiltonian virus. A ontradition. Consider now the

(7)

treated). Byenumeratingseveral ases resultingfrom the relative positions

of thearsin H,sets of disjointdireted pathsoveringV(H)donotverify

the onditionin order that(H;T +

;T )is ahamiltonian virus. Forthease

jT +

j =jT j =1( orjT +

j =jT j =2), wean seethe existene ofsets of

disjointdireted pathsoveringV(H)thatdonotverifytheonditionstobe

ahamiltonianvirus. ByTheorem2,Dis ahamiltonianvirus-freedigraphof

order 3.

InTheorem4,weneedthefollowingdenition:

Denition 1. [1℄A1-onnetedvirusisaloalongurationthat,ifpresent

in adigraph,forbidsthisdigraphtobe1-onneted. LetH =(V(H);E(H))

beaproperinduedsubdigraph ofagivendigraphD=(V(D);E(D)): A

3-uple (H;T +

;T ); whereT +

=fx2V(H):d +

H (x)=d

+

D

(x)gandT =fx2

V(H):d

H (x)=d

D

(x)g,isa1-onnetedvirus ifandonlyifV(H)=T +

or

V(H)=T .

Theorem4. A hamiltonian virus-freedigraphis2-onneted.

Proof. Let D = (V(D);E(D)) be hamiltonian virus-free. Letus reason ab

absurdo. Let (H;T +

;T ) bea1-onneted virus presentin D x forsome

x2V(D)withV(H)=T :TheaseV(H)=T +

istreatedinasimilarway.

Let y 2 V(D x)nV(H): By Remark 2, there exist vertex disjoint paths

P

1 ;:::;P

r

overingD y suh that eah oneof them makesairuit with

y. Sine V(H) = T then for eah P

j = x

1

j :::x

q(j)

j

we have x 1

j = 2 V(H):

Moreover, for eah x t

i

2 V(H)\V(P

i

)wehave x t 1

i

2 V(H). Hene x 1

i 2

V(H):Aontradition.

3 Hamiltonian virus-free digraph families

In this setion we desribe non-hamiltonian and hamiltonian virus-free

di-graph families. There exist non-hamiltonian digraph families with

hamilto-nian viruses. This fat hasallowedto deriveproblems and onjeturesthat

arepresentedanddisussedin thissetion.

Theorem5. Balaned bipartite digraphs are hamiltonian ifandonly if they

arehamiltonian virus-free.

Proof. LetD =(X[Y;E(D))beahamiltonianbalaned bipartitedigraph.

(8)

Remark2,D xhasaoveringbyvertexdisjointpathsP

1 ;:::;P

r

suhthat

eahoneofthemmakesairuitwithx. LetC

i

(1ir)betheseiruits.

Sine D isabalaned bipartitedigraph wehavejV(C

i

)j=2n

i

(the iruits

haveevenlength),jXj=n

1 +n

2

1++n

r

1andjYj=n

1 +n

2

++n

r .

ThereforeD isnotbalaned. Aontradition.

Thenextremarkfollowsdiretly fromTheorem77.

Remark 3. There are no non-hamiltonian and hamiltonian virus-free

di-graphs with minimum = 2 and nodes = 5. Notie that digraph D

5 has

hamiltonianvirus.

Theonlynon-hamiltonianand hamiltonianvirus-freedigraphwith

mini-mum=3and nodes=7isD

6 .

A non-hamiltoniandigraph withnodes =2minimum+19has

hamil-tonian viruses. By Theorem 77 the only families of digraphs that are

non-hamiltonian and where nodes =2minimum+19 holds, are D

7 and D

8 .

These familieshaveviruses.

Proposition1. A hamiltonianvirus-freedigraph withnodes5is

hamilto-nian.

Proof. Let D be a hamiltonian virus-free digraph; then minimum 2. If

nodes 4 then, by R

11

, D is hamiltonian. The ase nodes = 5 follows

diretlyfromRemark3.

Proposition2. Ahamiltonianvirus-freedigraphwithminimum=2is

trae-able orhamiltonian.

Proof. Sine minimum =2, there exists x 2 V(D) suh that d +

(x) = 2or

d (x)=2. ThenbyRemark 2,thereexist atmosttwovertexdisjointpaths

overing D x, say P

i = x

1

i x

2

i :::x

r(i)

i

(1 i 2), suh that eah one of

themmakesairuitwithx. IfthereisonlyonepaththenD ishamiltonian,

otherwisethepathx 1

1 x

2

1 :::x

r(1)

1 xx

1

2 x

2

2 :::x

r(2)

2

makesD traeable.

Proposition3. A hamiltonian virus-free antisymmetri digraph with nodes

=6,7or8ishamiltonian ortraeable. Moreover, the onlyhamiltonian

non-hamiltonian virus-freeantisymmetridigraphwith nodes=7isdigraph EX.

Proof. Let D beahamiltonian virus-freedigraph. Then minimum 2 and

woodall 4. By Theorem 65, if nodes = 6 then D is hamiltonian. For

(9)

Thefollowingonjetureshould betrue:

Conjeture 1. A hamiltonian virus-free antisymmetri digraph is

hamilto-nian ortraeable.

NotiethatdigraphEX isnon-hamiltonian, butitistraeable.

Thefollowingonjetureshouldbetrue:

Conjeture 2. Ahamiltonianvirus-freeantisymmetrir diregulardigraph

with r3andnodes4r+1ishamiltonian.

We an formulate the following remarks for Conjeture 2: ByTheorem

64,theonjetureforr=3istruewhennodes8. Forase9nodes13

the hypothesis hamiltonianvirus-free perhapsanbeuseful. Notie that by

Theorem78,theonjetureistruefornodes=2r+1. Theonjetureistrue

from Theorem66forr=5andnodes15.

Problem 1. LetD bean antisymmetriandhamiltonian virus-freedigraph.

Findthegreatestpositiveintegerx suhthatwhenarsnodes(nodes-1)/2 x

then D ishamiltonian.

ByTheorem4andTheorem67wehavex2. MoreoverthedigraphD

20

showsthatTheorem67isthebest possible. NotiethatD

20

hashamiltonian

viruses.

Problem 2. LetDbeak-onnetedandhamiltonianvirus-freedigraph. Find

the greatestintegerxsuhthatwhenalpha0k+x thenD ontainsa

(1,1)-fator.

ByTheorem 51wehave that x 0. Moreoverdigraph D

20

showsthat

this theoremisthebestpossible. Notiethat D

20

hashamiltonianviruses.

3.1 Hamiltonian virus-free hypohamiltonian digraphs

This setion is devoted to study hypohamiltonianhamiltonian virus-free

di-graphsand those that havehamiltonian viruses. Themethods, forbuilding

hypohamiltonian digraphs, established in [9℄ and [4℄ are given. Some

on-jetures related to hamiltonianvirus-free and hypohamiltoniandigraphs are

disussed.

AdigraphDishypohamiltonianifithasnohamiltonianiruitsbutevery

(10)

Conjeture 3. Everyhamiltonian virus-freenon-hamiltoniandigraphis

hy-pohamiltonian.

Ortheweakerone:

Conjeture 4. Every non-hamiltonian vertex-transitive hamiltonian

virus-freedigraph ishypohamiltonian.

Conjeture 5. Everyhypohamiltonian digraphishamiltonian virus-free.

Notie that digraph D

6

is hypohamiltonian and hamiltonian virus-free.

In[9℄Thomassengivesamethod forobtaininghypohamiltoniandigraphsby

formingtheartesianprodutofyles. Wegivehereashortsummaryofhis

results,in ordertogivesomeremarksonConjetures3,4,5.

ReallthatifD

1 andD

2

aredigraphsthenitsartesianprodutD

1 D

2

isthedigraphwithvertexsetV(D

1

this notationThomassen givesthefollowingtheorems:

Theorem6 ([9℄). Foreahk3;m2;C

k C

mk 1

isahypohamiltonian

antisymmetri digraph. Moreover, C

3 C

6k +4

is hypohamiltonian for eah

k0.

Theorem7 ([9℄). There isno hypohamiltonian digraph with fewer thansix

verties, andfor eah odd m3,C

2 C

m

is ahypohamiltonian digraph.

Remark 4. ThehypohamiltoniandigraphsC

3 C

6k +4

withk0(Theorem

6)andthehypohamiltoniandigraphsgiveninTheorem7verifyConjeture5.

Howeverthedigraph C

4 C

11

, i.e., k=4and m=3in Theorem 6,refutes

Conjeture5. Wehaveprovedthattheonlynon-hamiltonianvertex-transitive

digraph whih is also hamiltonian virus-freeof order 6, is the

hypohamilto-nian digraph C

2 C

3

: Whih is in favorof Conjeture 4. Nevertheless the

Conjeture 4is false, the digraphEX isnon-hamiltonian, vertex-transitive,

hamiltonianvirus-freeandnothypohamiltonian.

In[4℄Fouquet andJolivetgivethefollowingtheoremforobtaining

hypo-hamiltoniandigraphs.

sribedbelowishypohamiltonian.

(11)

Forn=2p+1andp3: V(F

takenmodulo 2p:

For n = 2p and p 4, F

n

is obtained from F

2p 1

replaing the ar

x

andaddingthefollowingars: x

2p 4

:Eah index istakenmodulo 2p 2:

InthenexttheoremletC =x

beairuit. Wedenote by

C(x

i ;x

j

)theinduedpathofC beginningat x

i

andendingatx

j .

Theorem9. Foreah n8;F

n

ishamiltonian virus-free.

Proof. WefollowRemark1andRemark2. Ineahstepoftheproof,weshow

the pathsP

i

). Weonsider twoases:

Case 1n=2p+1andp4:

Remark 5. For n8,thehypohamiltoniandigraphsF

n

(12)

4 Conlusion

It is well known that the problem to deide when a digraphis hamiltonian

is NP-omplete[3℄. A \yes"answertothehamiltoniity problemforagiven

digraph an be veried by heking in polynomial time that a sequene of

vertiesgivenbyanoraleisahamiltonianiruit.Inaseofnon-hamiltonian

digraphs, asstatedin [7℄ pages28, 29, there isno known wayof verifyinga

\yes"answertotheomplementary problemof deidingifadigraphis

non-hamiltonian. A solutionto this problem is to provide a hamiltonian virus,

whose presene in the digraph an also be heked in polynomial time. In

aseof thenon-hamiltonianhamiltonian virus-freedigraphs,theymusthold

the partiularstruture givenin Remark 2. Thevirus notionhasbeenused

in randomgenerationofdigraphswithoutertainproperties[8℄.

Wehavebuilt aninterativesupport toolalled GRAPHVIRUS [5℄that

allowsthegraphial edition ofhamiltonian viruses and theveriationthat

agivenstrutureisahamiltonianvirus. GRAPHVIRUSanalsobeusedto

deriveaproedure fordeidingwhether agiven digraphisnon-hamiltonian.

Thisproedureisofthesameomplexityoftheproblemofdeidingifagiven

digraph ishamiltonian, butthe interestoftheproedure isthefat ofusing

aloalstruture.

Finally,thetheoretiinterestoftheresultspresentedhereistheirrelation

with the extension of known suÆient onditionswith the new hamiltonian

virus-freeonditionfortheexisteneofhamiltonianiruits.

Referenes

[1℄ M. R. Brito, W. Fernandez de la Vega, O. Meza, O. Ordaz, Viruses in

GraphsandDigraphs,VishwaInternationalJournalofGraphTheory,

Vol-ume2,Number1(1993),35{55.

[2℄ Ch.Delorme,O.Ordaz,D.Quiroz,Tools for studyingpaths andylesin

digraphs, Networks31(1998),125{148.

[3℄ Ch.H.Papadimitriou,K.Steiglitz,Combinatorialoptimization,

Prentie-Hall,NewJersey,1982.

[4℄ J.L.Fouquet,J.L.Jolivet,Grapheshypohamiltoniensorientes,Colloques

internationauxC.N.R.S.N o

(13)

[5℄ L.Freyss,O.Ordaz,D.Quiroz,J.Yepez,O.Meza,GRAPHVIRUS:Una

herramientaparaeltratamientodefallasenredes,ProeedingsdelaXXIII

ConfereniaLatinoameriana de InformatiaPanel'97, Valparaiso,Chile

(1997),243{252.

[6℄ L. Freyss, O. Ordaz, D. Quiroz, A method for identifying hamiltonian

viruses, Proeedings de la XXIII Conferenia Latinoameriana de

In-formatiaPanel'97,Valparaiso,Chile (1997),231{242.

[7℄ M. R. Garey, D. S. Johnson, Computers and Intratability - A guide to

theTheoryofNP-Completeness,W.H.FreemanandCompany,NewYork,

1979.

[8℄ F.Losavio,L.E. Marquez,O.Meza, O.Ordaz, La Generation Aleatoire

de Digraphes dans l'Environnement AMDI, Tehniques et Siene

Infor-matique,Vol.10,No6,(1991)437{446.

[9℄ C.Thomassen,Hypohamiltoniangraphsanddigraphs,inTheoryand

Ap-pliations of Graphs, Leture Notes in Mathematis, Vol. 642, eds. Y.

Referensi

Dokumen terkait

(For a topological space in general this is a preorder, not necessarily anti- symmetric, but for a T0 space, as also for a locale, it is a partial order. A space is T1 iff

As we previously said, in order to apply the twist fixed point theorem to prove the existence of periodic solutions to planar Hamiltonian systems, Birkhoff tried to replace

and Yang C.-C., Some further results on the zeros and growths of entire solutions of second order linear differential equations , Kodai Math. and Yang C.-C., On the zeros

It is also shown that the maximum number of Hamiltonian cycles in (sparse) graphs for which the number of edges is less than 1.5 times the number of vertices, only depends on the

The approach taken here is to use an asymptotic approximation for the &amp; undershoot ' , the amount by which stock has fallen below the re-order level by the time an order is

The &#34; rm is now free to choose manufacturing capacity, while (correctly) taking into account the di ! usion phenomenon. 3 shows that optimal initial manufacturing capacity

The issue is the relationship between the over-dispersed Poisson (ODP) model of Renshaw and Verrall (1998), the distribution-free stochastic model (DFCL) of Mack (1993) and

The main new element of our results is a high order exponential asymptotic expansion which extends in a unified way both the classical Cramér–Lundberg approximation and the