MATHEMATISCHES FORSCHUNGSINSTITUT OBERWOLFACH

Tag u n 9 s b e r i c ht ")1/1984

Graphentheorie

8.7. bis 14.7.1984

Die Tagung fand unter der Leitung von' Herrn' G. Ringel (Santa Cruz, California) und Herrn W. Mader (Hannover) statt.

Mehr als 40 Teilnehmer aus 13 Ländern (Australien, 'VR " China, CSSR, Dänemark, England, Frankreich, ISI;'ael,.·· Jugoslawien, Kanada, Niederlande, Ungarn, USA) haben über ihre n~uesten Ergebnisse in verschiedenen Gebieten der Graphentheorie b~richtet. In einer von Herrn P. Er~' dös geleiteten Problemsitzung wurden offene Probleme' vorgestellt und diskutiert.

Im besonderen wurden Resultate aus den folgenden Teil- ~ gebieten der Graphentheorie vorgetragen:

Färbungsprobleme, (Kanten- bzw. Knotenfärbunge~/·Larid~· ka.rten) , zusammennangsprobleme, Unendliche. Graphen, Einbettungen.von Graphen in verschiedene Fiächen, (Mini­ malbasen, Kantenkreuzungen), Partitionen (Turan-Graphen, Ramsey-Theorie usw.), . ~. .. '\ Wege und Kreise in Graphen.

Darüber hinaus wurden graphentheoretische Methoden be-

© nutzt im Zusammenhang mit Matroiden, Netzwerken, Opti­ mierungsproblemen,' physikalischen Problemen, Lateini­ schen Rechtecken.

Als spezielle Ergebnisse seien noch die Fortschritte im "Oberwolfacher Problem" erwähnt sowie die Lösung der seit 1890 offenen Imperium-Vermutung von Heawood. Die Grundidee zu dieser Lösung ist in der abgebildeten Land­ karte mit 19 paarweise benachbarten, aus jeweils 4 ge­ trennten Ländern bestehenden Staaten angedeutet.

© - 3 -

vortragsauszüge-

R. AHARONI: Matchings in infinite graphs

A criterion is present~d for a graph· at" any cardinality to passess a perfeet matc~ing.

B. ALSPACH: Same new results on. the Oberwolfach problem

Let F(11,12, •.• ,lr) IG denate an isomorphi~ factorization of G- into 2-factors each of which is compased of cycles af lengths l1,12, ••• ,lr. If a~l the li'5 are the same value 1, write F(l»)G instead. The results mentioned in this talk are'the following. Thm.1. F(2m) IK2~-I for all positive -integers r~1 and ~2 where I 'de'notes a·. 1-factor of K2~. Thm.2 •. If ~i~4 and even for each i, r th~n F(11.,12' .•• 'lr)IK4m,4~ for every m'~ 1 with L 1. Sm. i=1 1

Thm.·3. odd and n odd, then F .(d~_) 1K If _F(s) IKn , ~ dn for all odd d. ~(s) aso long as s i5 odd, s=9.~m~d3) and Cor.4. IKn n is an pdd multiple of s. The above is joint work .with Roland.Häggkvist.

T. ANDREAE : AGame of Cops and Robbers' There are two players c = cop player and r = robber player. First c places s cops at same of the vertices of a finite connected undirected graph, G. (Twa o~ mar~ c~~~ may be on the same vertex. ) Then' r "places ä. robber at some ·vertex. Thereafter the players move alternatively·.·· A move of' c corlsists of moving"some'of the cops along' edges to:adjacent vertice5. Similarly; a move of·r i5 defined. c' 'wins if he catches the robber, r wins if he'· avoids this forever". Let· c (Gl be the minimal number of cops that are sufficient to catch the robber~ AIGNER and FROMME proved that c(G)~3 if G' is planar, and that,

© - 4 - in general, c(G) can be arbitrarily high. Here it is shown that, for each finite graph H, there is a minimal a(H)EN such that c(G)~a(H) if H is not a subcontrac­ a(Kn)~(n-1) n~4, tion of G. Further, (n-3) for O(KS) 3, O(R ,3) 3, a(K 2, a(Wn)~rn/31+1 (W wheel = 3 = S) = n = with n rim vertices, Kn(K~) = with n vertices (minus an e~ge». Other results: 1. QUILLIOT showed c(G)~2n+3 if G has genus n,~. an algorithmic characterization of the graphs with c(G) = 1 is due to QUILLlOT and NOWAKOWSKI/WINKLER, 3. c(G)~2 if !V(G) 1~10, G # Petersen Graph.

I. BEN-ARROYO HARTMAN: Path Partitions and Packs in Digraphs

A path partititon P {P , ••• ,P } of a digraph is a = 1 m partition of its vertex set inta disjoint directed paths. A path partiti~n P is k-optimal if I: mln{ IP',I ,k} is as P.EP· l. l. small as possible; IPil is the number of vertices in k Pi. A partial k-colouring C is a collection of k dis-, joint independent sets. Berge conjectured the following: Let G be a directed graph and let k be a positive integer. Then for every k-optimal path partition. P there k exists a partial k-colouring C such that each path Pi in P meets min{IPil,k} different colour classes k of c • We report on recent progress on the problem and on a dual analogue, interchanging the roles of paths and colour classes.

R. BODENDIEK: Uber ebene Graphen

Ein endlicher, ungerichteter, schlichter und in die Kugel Fo einbettbarer Graph" G = (E,K) ist dadurch ausgezeichnet, daß er höchstens einmal gesä~tigt werden kann und daß der Grenzgraph G(L) für jedes Land L der ebenen Landkarte dag~gen (G,Fo-G) ein Husimibaum ist. Ist F * Fo eine orientierbare oder nichtorientierbare Fläche, so lassen sich stets 'in F e1nbettbare Graphen G finden, die min­ destens zweimal gesättigt werden können oder bei denen

© - 5 -

Länder in der Landkarte (G,F-G) existieren, deren Grenz­ graphen keine Husimibäume·sind. Betrachtet man dagegen die Spindelfläehen o(Cn ), bei.denen der äußere Graph G ein (n~3) jeder'der'i~eren Kreis Cn ist und Graphen -G" 'G2 , ~o Spindel~ ..• , Go ein ,K 2 , kann man zeigen, daß diese flächen für sogenannte normale Graphen, die sich überdies

in o(Cn ) so einbetten lassen, daß die' n singulären Punkte von ~(~n) Ecken des eingebetteten Graphen'sind; die gleiche Sättigungs- und Landkarteneigenschaft wie Fa j~doch besitzen. Im Gegensatz zu X(Fo ) = 4 ist die chro-' matische Zahl X(O(~n» für jedes n.?: 3 gleich 5.

A. BONDY: Parity Theorems for Paths and Cycles

, , , This is areport of joint work with F.Y.' Halberstam. A gra~h is ~ (resp. odd) if every vertex has even (r~s~. od~) d~­ gree. Using a technique of A.G. Thomason, we prove the fol-' 'lowing results. Theorem 1. Le,t , Pi (u) ,be the number of paths }?f length .i with initial vertex u. Then (a~ if G is even, .~~(~) is even (i~1), (~) if G 1s odd, Pi(u) is even~(i,?:2). Theorem ,2. Let Pi be the n~er of paths of leng~~ i . . - Then (a) if G is even and bipartite, Pi. is eve~ (i~1), (h) if G 1s odd and bipartite, Pi is even (i~3). Theorem 3. Let v be the.number 'of vertices' and p the number of paths. Then (a) if G 1s even, p = v(mod2), , 1 (b) if G 1s odd, P = 2v(mod2) • Theorem 1 has ari application to the question of exlstence" of generaliz~d friendship graphs (th?se graphs with the' '" • property that, for a given k,?:3, there is a~unique (u,v)-pa,th of length k for all pairs u,v of d1s.tinct vertices)':' A. Kotzig has conjectured that no such graphexists. Theorem 1 (a) imp~1es that each such graph i~ ?f o~~ order.

A. BaUCHET: Isotropie systems. Recognizing circle grap~s

Let V be a finit~ set. We consider a, 2-dim.ension~l·.yec::tor space K over GF(2) with the non-null alternate bilinear " .. ", . '. 'V'; form (x,y) ....xy and the symplectic structure induced on K

© -.6 -

by the alternate bilinear form (a,ß)~rVEVa(v)ß(v). An iso­ tropie system tL,V) is defined by a totally isotropie sub­ V spaee cL ~ K of .dimension lVI. Isotropie systems ean be assoeiated to 4-regular graphs ­ ealled graphie systems - and also to pairs of dual binary matroids. In.ge?eral isotropie systems unify some proper­ ties of 4-regular graphs and binary matroids. The application which we develop here comes from the facts that eaeh isotropie system ean be identified to a simple graph defined up to Ioeal eompiementations and t~at ~he cireie graphs (defined by assQeiating vertiees to chords of • a cireie and joining vertices iff the corresponding chords intersect) correspond so to graphie systems. A theory of 3-connectivity for isotropie systems similar to Tutte's theory for matroids allows to derive an efficient algorithm for recognizing circle graphs.

P. A. CATLIN: Homomorphisms into add eycles

A harnomorphis~ 8: G ~ H is a function fram a graph - G into a graph H such that x - y irnplies Sex) ~ 0(y) , where denotes adjaceney. We eonsider hornomorphism~ as ä generalization of the coneept of vertex coloririg (the case with H eomplete). There is a construetive characterization of the edge-minimal series-parallel graphs having no homomorphism inta eS .

For the ease H = C + ' there are variaus. recent results, 2k 1 including a sufficient condition for a homomorphism onto an odd cycle to ~e unique. • G. CHEN: Spanning balanced bipartite subgraphs of regular graphs - On R. HÄGGKVIST's Conjecture

This paper points out a counter-example' of R. Häggkvist's conjecture: "Let .G be an rn- with an even number of vertices, then G contains a spanning balaneed bipartite graph- B where each vertex has degree ~[~]."

After modifying the original conjecture, we have proved. that for m ~ 3 the modified conjeeture is true.

© - 7 -

D. CVETKOVIC: Inte'ractive prograrnming system "Graph" ­ an expert system for graph 'Lheory

Interactive programming system "Graph" has recently been implement~d at Faculty of Electrical Engine~ring, University of Belgrade. System "Graph" contain~ a c~mputarizeq.grap~ theory biblio~raphy, a subsystem for ,performing several tasks on particular graphs, and a theorem prover. The purpose of the system is to support research in and applications. Userls manual, a protocol of a conversa­ tion with the system "Graph" and other material on the system will be.presented to interested persons. w. DEUBER: Toroidal graphs admit a 5-flow Tuttels conjecture: For every finite graph G without bridges there exists an orientation G of the edges and a flow (Kirchhoff flow) lP: E(.G) .... {1,2,3,4.1 . THEOREM (M. MÖLLER): Tuttels conjecture holds for graphs of genus 1 .

A. FRANK: Edge-disjoint paths in planar graphs'

The following theorem is presented. THEOREM.ln a planar graph G k pairs of terminals on the boundary are specified. Every node not on the boundary has even degre~. ~hen ther~ exist ,k' e~ge-disjoint'p~~hs in G connecting the corresponding'terminals iff

L'surplus (Ci) ~! ~amily ~ ~-J for eyery. C1 ,C2 ,•·• ,Cl.} of _. 1 IV I cuts, where q denotes the ~urnber of odd components in G-C -C .••C . 1 2 1 (The surplus of a cut C is the difference between the number of edges in C and the n~er 9~ terminal pai~s separated by c. A set X ~ V is odd i~ the n~er of edges le~ving X and the number of terminals in X have different parity.) . . '. Th~s ~~orem is ~ common general~zation of earlier results of Okamura-Seymour and of the author.

© - 8'-

C. GODSIL: Enumerating Latin Rectangles

Working jointly with Brendan McKay (Canberra), I have derived an asymptotic formula for the number of kxn 6 7 Latin rectangles, valid for k = o(n / ). The pr~vious best result was accurate only for k = o(n1/ 2 ). The calculation is based on an estimate for the number of ways of extending a kxn Latin rectangle R to a (k+1)xn rectangle. In fact this number is

Here G is a subgraph of Kn,n determined by R -and n n k r(G,K) ~ (_1)kp (G,k)K - k=1 where p(G,k) is the number of rnatchings in G with k edges. This leads to an estimate of the number of exten­ sions in terms of n,k of certain small subgraphs

... ) of G.

With some difficulty this estimate can be averaged over all kxn rectangles R, thi~ leading to the final result.

M. GRöTSCHEL: Eulerian SUbgraphs, Cuts and Certain . Binary Matroids

In this talk we study the convex hull of the incidence vectors of the cycles of a binary matroid. We prove that a descriptionof the facetsof this polytape can be ob­ tainded from a description of the facetsthat contain any vertex. A compl~te and nonredundant description of this polytope by linear equations and inequalities i8 given for those binary mat~oids with no F *, R or M(K )* 7 10 S minor. This implies a convenient characterization of the convex hull of the incidence vectors of Eulerian sub~ graphs of a graph and of the convex hull of the incidence vectors of the cuts in a graph not contractible to K • S

© - 9 -

I. GUTMAN: Graphie polynomials whose Zeros are Real .. It is weIl known that the eharacteristie polynomial

' 137-144.) Let the graph G have n vert~ees V 1 ,V2 '·· .. ,vn • Let A ,A , ••• ,A be no~-negative real' numbers • 1 2 n .' "'n (~) Then the polynomia1s

R. K. GUY: Outerthickness & Outercoarseness of GraJ?hs ._.

A graph is planar.if it can be imbedded iQ t~e plane (ar sphere) •. The eomplement of the graph in. such an imbed­ ding is a collection of open polygons, whose boundary eon­ sists of v~rtiees joined by edges. I~ there i~,an,i~~d­ ding with all the vertices of th.e .graJ?h qn the boundary. ~.f ..( a single polygon, the graph is said to be outerplanar. ~ .

ehar~cterization of ou't:erplana:r:. graphs. is tha.,t. ~hey.. d9.. ". ..'';' not contain a subgraph homeomorphic to K4 or :K.: 2 , 3 • ':.' The thickness (resp. coarseness) of a graph is the minimum (resp. maxi.mum) nUinber of' pl'anar (resp~' non-planar) graphs into whieh the edges of the graph can be parti­ tioned. Outerthiekness and otitercoarseness. ~re .corr·esP9.~~.:-~~··· dingly defined by edge-partitions into (non)outerplanar ~ •• p ~ ....,• ; ~ ,-. 7" .•• i k ~. ; graphs. We g.iv~ tp.e .outerthickness and. o~terco_~rse~es.s of the usual families of graphs: Kn , Km, n' and the·- ... ,'. . d-diminsional cube.

© - 10 -

E. GYöRI: Edge disjoint comp1ete subgraphs

One of the elassieal theorems in graph theory is Turan's theorem. We prove some theorems' about deeompositions of graphs into edge disjoint eomplete subgraphs whieh sh~w the stronger extremality of Turan graphs. E.g. we prove that any graph of n vertiees can be deeomposed into Kk'S and edges so that the s·um of the orders of these eomplete subgraphs is no more than the surn of the degrees of the k-1 partite Turan graph, providing that k is at least, 4 . We also give esti­ mates for the number of the edge disjoint triangles in a graph of n vertiees with m edges.

A. HAJNAL (jointly with P. Erdös): On 3-partitions of a set

We give two results relevant· to the Ramsey funetions R (n,n) and R (n,n,n). It is weIl known that 3 3 c n 2 c,n 2 2? < R (n,n) < 2 3 for sorne eo,e, > o. The'next theorem'says that the lower boun~ can not be improved using "justly distributed" parti­ tions ... ~.> ~ Theorem 1.' 3 3c a > 0 \J-m V(Ko,K,) ([rn],3 =', Ko U·K, :::;0 3H c: In 3i < 2

I H I ~ Ca' Vlogm "I [H] 3 n i< i I > a (I ~ I))., 2 n logn Theorem 2. R (n,n,n) -> 210g.1ogn for n > n • 3 o We conjecture that' Theorem 1 extends to' every a < i. For history and ear1ier results see "Combinatorial Set Theory" by Erdös, Hajnal, Mate, and Rado.

© - 11 -

H. HARBORTH: Multiple crossings in drawings of graphs

Realizations of a graph in the plane are considered where two lines have at most one point in common, either an end­ point or a crossinq. For graphs with 2m vertices at most rn-feld crossings are possible. It is .proved, that the maxi­ mum number ?f m-fold crossings is 2 for m = 3 and m = 4, and at least 2 in general.

A.J.W. HILTON: Edge-colouring of graphs

.~hi~. 1s joint with myself and A.G. Chetwy.nd. Let x' (G) and. 6(G) denote the ~dge-chro~atic number and maximum degree of a simple graph G. If IE(G) I. > A(G)lIV~~) IJ., tpen ca~l G an, overfull graph. ~t i~ well-known that if G ~s an,ove.rfull graph then ~'(G) = .6(G)+1., Theorem 1. Let G have r vertices of maximum.. degree, let 7 IV(G)I' 2n or 2n+1 and l~t a(G) ~ n + 2' r - 3. If Xl (G) A(G)+1 than it follows that either G 15 overfull er G .has an edge-cut. ~ w~th 181 < r - 2 . such that

~(G) G ...... 8 =:= G" UG2' where G 1 n G2 .= cf>, .l~ (G 1 ) and G1 i5 overfull. Theorem 2. Let I.V(G) I be even and let G be regular of 6 degree d:'(G), satisfying d(G) ~"7 )V(G) 1.. Then .x l (G,.) Ö (G) •

L. LOVA8Z: Independent 's'ets in claw-free graphs

A claw. in a graph 15 .an .induced .K~1,"3. The class of. claw-free graphs. includes all lihe-graphs',' and, many non-l~ne graphs, e.g.

The proble~ of, find~ng a ~~ximum-i~dependent set of' points in a claw-fre~ gr~~h includes the ..~el+~solved matching prob~.em. Polynomi~l.time algorithm~__ ~or this problem w~re .givel?- by '. Sbihi and Minty. In this t~~k it 1s s~o~ that every .claw-~ree graph arises from a line-graph by gluing on pieces with inde-

© - 12 -

pendence number ~ 2 in a well-described way. This gives use to an algorithm which reduces the independence number problem for claw-free graphs to a matching problem.

W~ MADER: On minimally n-connected digraphs

Für einen minimal n-fach zusammenhängenden, gerichteten Gra­ phen D sei 0 der von den Kanten (x,y) mit y+(x:O) > n und 0 Y-(YiD) > n erzeugte Teilgraph, wobei y+(x:O) bzw. y-(x;D) den AUßengrad bzw. Innengrad von x in D bedeute. Es wird ge- 4It zeigt, daß 0 keinen alternierenden .Zyklus enthält. Dies ist 0 äquivalent dazu, daß ein 0 zugeordneter paarer-Graph keinen 0 Kreis enthält. Hieraus ergeben sich ähnliche Resultate wie im ungerichteten Fall. (1) Für jeden endlichen, minimal n-fach zusammenhängenden, gerichteten Graphen 0 -gelten !{XEE(D): "Y+ (x; D) =n} I ~ n lind I{xEE (0) : y + (x; D) =n} I + !{xEE (D) : y - (X-i D) =n} I• ~ ~~~1 2101 + ~n-1 • (2) Für jeden endlichen, minimal n-f~ch zusammenhängenden, gerichteten Graphen 0 gilt für die Kanten- zahl 1101 I ~ 2nlDI - n(n+1) und im Falle IDI > 4n + 5 so- gar 1101 I ~ 2n (IOI-n), wobei die D charakterisiert werden, für welche das Gleichheits~eichen gilt.

I. MENGERSEN: Eine Abschätzung für die Ramseyzahl r(KS-x)

Unter der Ramseyzahl r(G) eines Graphen G versteht man die kleinste natürliche Zahl- p, so daß-bei jeder 2-Färbung

der Kanten des Kp ein einfarbiger Teilgraph G vorkommt. ~ Bislang ist r(Kn ) nur für n 4 exakt bekannt, für ~ ~ r(KS ) weiß man nur 42 r(KS) 55 . Als eine Art An­ näherung an r(KS ) ist r(K5-x) von Interesse (KS-X ent­ ~iner steht aus KS durch Entfernen Kante). Die beste bis­ her bekannte obere Schranke für r(K5-x), 24, wird zu 23 verbessert. (Die beste bisher be~annte untere Schranke ist 21.)

H. MEYNIEL: Partitions of Digraphs into Paths or Circuits

An upper bound on the minimum number Of paths or circuits in a partition of the ares of a digraph arid same conjec­ tutes related to this topic are given.

© - 13 -

R. H. MÖHRING: Two Theorems on Graph Substitution

Suppose we want t~ represent the maximum clique weight wG(xf of G as a function of the'node weights x = (XVIVEV) in a decomposed way as [f1(xvlv~V1) wG(x) f , ••• ,fm(xvlve:vm)]

f,f".~.,fm where V" ... 'Vm form a partition of V, and are realvalued functions. 'Under certain conditions on.the 2 .?ijec~iy~ly f i (no f i may map a subset homeomorphic to R into Rl ) we show that G must then decompose according to graph substitution (= X-join) as

G = G'[G'lv1, ••• ,Glvm],

where GIVi is the subgraph of G induced by Vi and G' is the'quotient graph of G with respect to the partition {Vl,···,Vm}· Appl~g results on the asymptotic relative frequency of (subs~itu~ion-) indec~mposable partia~ orde~sl we .then ~~ow that almost all comparability graphs are uniquely parti~l~y, orderable (UPO), i.e.

lim # UPO comp. graphs with n points = # comp. graphs with n points

As a conseque~ce, the number of compa~ability.graphs.is asymtotically equal to half the number' of.parti~l orde~~.

H. M. MULDER: Distance-hereditary Graphs

Distance-hereditary graphs are connected g~aphs. ~I.l ~hich,. ," all induced graphs are i~ometric (i.e. all'induc~d paths:. are geodesics). Examples of such graphs are provided ~y ,-. complete, multipartite graphs and ptolemaic graphs. Every finite distaJ)ce-hereditary graph can,be, obtained . . from K, by iterating the following two. opera~ion~: , adding pendant vertices and vert~x splittings. Using'. this~.· result characte~izations of (infinit~) distance-heredit~ry graphs can be deduced: in terms of the distance function d

© - 14 -

(involving some four-point-condition), or in terms of the interval funetion I of the graph, or via forbidden iso­ metrie subgraphs. Related results on ptolemaie graphs and parity graphs are given.

J. PACH: On universal graphs

A elass ~ of infinite grap~s is said to have a univeral element Go E 9f if G c Go holds for every GE: CJ · Theorem 1. The class of countable planar graphs does not have a universal element. Theorem 2. (P. Komjath & J.P.) Assume GCH. Let < a ~ ß ~ Y be cardinals, a < w ~ Y Then the elass lJy(Ka,ß) of all Ka,e-free graphs on y vertiees has. a universal element iff

(i) y > w or

(ii) y = w a = and ß ~ 3 •

The special.eases (a=y>w,ß=1) and (a=ß=2,y=w) of Theorem 2 were p!oved by Shelah (1973) and' Hajnal-Pach (1981), Theorem answers a question of Ulam.

K.-P. PODEWSKI: Critical families

Let F = (F(i) liEI) beafamily of sets. F is called critieal if F has an injeetive ehoieefunction and f[1] = U{F(i) liEI} for every injective choieefunction. Critical families can be defined by transfinite recursion on Cl as foliows. If a = .Ua , then F E Ca iff there is a chain (I ) ß

© - 15 -

G. RINGEL: Beweis der, ebenen I-mperium-Vermutung von,Heawood

Es wird bewiesen, daß in der Ebene eine Landkarte existiert mit Gm paarweise benachbarten Staaten, wobei jeder Staat aus m getrennten Ländern besteht (m~2). Dies war bisher nur für' m=2 von Heawood "'890 und für m=3 oder 4 von Taylor 1980 be­ wiesen.

Y. RODITTY: Packing and covering of complete graphs by trees

It is shown that: [~~~=~n gl~:~l, p(n,Tk } = and C(n,Tk } = n > 9 , where T k is _.any tree of order k ~ 6,,' P(n,G) the maximal number of pairwise,edge disjoint graphs G in the complete graph K n , and C(n,G) the minimum number of graphs G whose union is Kn • ([xl de­ notes 'the largest integer not exceeding x and {x} the least integer,not less than x.) It was' as~ed ~he ~ollowing: I) Is it true that for each tree T of order k > 6: k P (n,T ) = (n-1)] and C (n, ,T ) {n (n-') 1 k [n2(k-l) k 2(k-1)J (no som~constant to be determined) .

11) 15' it true that for' each integer' k > 6, there,exist integers' .m,'n" such that' if (k-1) I mn then K 'is m,n deeomposed inta isomorphie eopies of each tree T of' k order k? Ex: For k G I K4 ,5 is decomposed into each tree of order 6.

V. RÖDL (joint' result with Peter Frankl): 'Oensity theorems for'graphs and hypergraphs

Let G: (n.,e.), i = 1,2, ... , n~ ~ m be a sequence of 1. ~ 1. .... r~uniform edges~ hypergraphs of n i vertiees and e i The density of this sequence is a, '0 ~ a < 1, if a is the largest'real number for whieh there"is'a subgraph G~, G~ ~.'X) (mi,e'i) of '(ni,ei ) for whieh mi and lim e!/(m/) = a • The theorem af Erdös, Stone and Simonovits 1 .

© - 16 -

states that for r =_ 2 the only possible values of the densities are 1 and - .! where t t = 1 ,2, ... It was conjectured by P. Erdös that for r > 3 the set Dr c [0,1] of possible densities forms a weIl ordered set. We disprove this conjecture and show the following

Theorem: For every E > 0, r ~ 3 there exists l(r) such that for any 1 > l(r)

1 - 1 < d < 1 - 1 + E i r - 1 rr-1 for some d EDr

I. SCHÖNHEIM: Application of small Turan numbers

Let f: E(K ) ~ E(K ). The K A is caiied free if n n 4 fee) E E(A) for eE E(A). Theorem 1. If fee) * ethen the smailest number enforcing a free K4 for every f is 10. Theorem 2. The smallest number enforcing either an edgewise. fixed triangle or a free triangle is 13.

A. SCHRIJVER: t-perfect graphs

An undirected graph G = (V,E) is called t-perfect if the coclique polytope (= convex hull of the characteristic vectors of cocliques (independent sets» is determined by the linear inequal~ties

(vEV) ,

x +x < 1 (vw E E) t w = veC Xv ~ l~lcd (C odd circuit) . If a graph is t-perfect, we can find a maximum weighted coclique in polynomial time, 'with linear programming s~ow methods. The graph K4 is not t-perfect. We that a grapp is t-perfect if it does not c~ntain a homeomorph

of K4 as a subgraph in which all triangles of K4 have become odd circuits. This extends results of Boulala, Fon~upt and Uhry. We also discuss a decomposition theorem

for graphs without these "odd K4 's".

© - 17 -

H. SCHUMACHER: Zur Struktur primer 1-optimaler Graphen

Ein l-einbettbarer Graph mit 00 Ecken und 01 Kanten heißt prim l-opt~al, wenn 01 ~ 40.0 - 8 und die Zusammen­ hangszahl z(G) = 6 ist. Ist r* die Menge aller primen I-optimalen Graphen, so ist die Bestimmung der chroma~ tischen Zahl x(r*) gleichbedeutend mit der Lösung des von G. Ringel angegebenen 6-Farbenproblems für Landkarten.

Auf r* läßt sich eine Ordnungsrelation >R definieren, deren Minimalbasis ~'M (>R' r~) die Menge der. Graph~n .. 2 * ,m€ N'{1,2} ist, wobei C e~n Kreis der Länge C2m 2m' 2m. ist, in .~em zusätzlich je zwei Ecken mit dem Abstand 2 durch eine Kante verbunden sind. Für Graphe~ GE r*'M(>R,r*) gilt: 1. Ist keine .Kante von G, so ist k bei.j~der I-Ein­ bettung von G kreuzungsfrei od~r k wird bei jeder l-Einbettung in G von einer anderen Kante von G ge-· kreuzt. 2. Löscht man in G alle sich kreuzenden Kanten, so spa~nen ~ie Ecken mit minima~em Eckengrad in dem so erhaltenen Graphen V(G) einen Wald au~, dessen Kompo­ nenten Wege oder Dre1ersterne sind.

K. STEFFENS: Matchings and maximal tight sets

A tight set T c V of a graph (V,E) is a matchable set such that s (F) = T for all matchings F o~' T" (s (F) = = U {{x,y} I {x,y)€F}). ~ Every graph has a maximal tight . . set and every maximal rnatchable set" 1s a maximal tight set. But 1f G =. (V~E) " 15· countable, then every maximai tight set 1s also ~ maximal rnatchable set o~ vertices. Therefore every cöuntable ~raph has a maxi~al matchable" set of vertice~s. D.efine by ?l(T) ='TU .{xEv'TIN(X)~} the ctosure cf T and by kernG = U{cl(T) .IT~V tight} the kernel of G. Then for every maximal tight sets 1','5 we have cl(T) = cl(s), so kernG 1s equal to the closure of any maximal tight ·set. One can now prove a decomp?sition theorem for the k~rnel of a gr~ph. If G is ~inite, th~n this theorem is the Edm6nds'-Gallai decomposition theorem' for matchings:

© - 18

B. TOFT: A S11rvey of n-l1nked Graphs

A graph 1s called n-linked if there for distinct vertices

x 1 , ..• ,xn 'Y1' ••• 'Yn exist n disjoint paths, the first from x 1 to Y1' the second from x 2 to Y2' etc. The study of n-linked grapns was initiated in 1967 independently by M. Watkins and R. Halin. Halin asked if a graph is n-linked . . under the assumption that it.has a sufficiently high con- nectivity h{n). The problem has four different versions (undirected or directed graphs; vertexdisjaint or edgedis­ joint paths) • The talk gives a survey of same af the problems and results . on Halin's problem and the more general related subgraph homeomorphism problem, among them: 1) In the undirected vertexdisjoint case the existence of h(n) proved by H.A. Jung and D.G. Larman & P. Mani, based on a result of w. Mader, 2) In the undirected edgedisjoint case a generalized version of Menger's theorem by K.E. Strange and B. Taft, 3) In the directed edgedisjoint case the com­ plete solution of Halin's problem by Edmonds' branching theorem. In the directed vertexdisjoint case Balin's problem 1s still open.

K. WAGNER: Graphen von Graphen und Spindelflächen

Es werden zu. jedem Graphen G eine Spindelfläche ~(G) und Graphen von Graphen G(G1'.•• 7Gn) definiert, wobei die G , •.•,G beliebige plättbare, zusammenhängende 1 n Graphen sind und die~e anstelle der Kanten. k 1 , •.•,kn von G in den beiden Endpunkten von k 1 r • .• ,kn jeweils angeheftet sind. Es gilt dann für die chromatische Zahl X(r> jeder Spindelfäche r = 6(G) der Satz: X (0) = 4 oder = 5 oder = x(G). Daraus folgt speziell, wenn (auch) G plättbar ist: X(0) = 4 oder = 5 und X(ö) = 5 G enthält einen Kreis. Weiter folgt, daß zu jeder natürlichen Zahl Nein 0 = O(G) existiert mit X (0) = 5 und X(G).:5. 4 für alle in 0- einbett­ baren Graphen G mit einer Eckenzahl IE(G)I < N• Von besonderem Interesse ist die Fläche ~1 ' die aus dem

© - 19 -

Torus anscha\llich durch "Zusanunenschnüren" eines Breiten­ kreises aut ~inen p~nkt (= singulärer p~nkt von Ö, ) entsteht. Ein Graph heiße minimaler Graph von ~1' wenn er nicht in 0 , einbettbar, aber nach Lö~chen oder Kon­ traktion jeder beliebigen Kante desselben stets in ~, einbettbar ist. Z.B. sind der K6 und der Petersensche Graph ~wei solche minimalen Graphen von 1)1

M. E. WATKINS: Infinite Paths Containing Only Shortest Paths, and How to Avoid Them

Hier werden nur zusamrnenhängede unendliche, lokalfinite Graphen G betrachtet. Ein 2-seitig (bzw. 1-seitig) un­ endlicher Weg A in G heißt Achse (bzw. Halbachse) , wenn ein kürzester Weg, der je zwei Ecken von A ver­ bindet, in A selbst enthalten ist. Nach einem bekannten Satz von D. König enthält G einen 1-seitig unendlichen Weg. Wir beweisen das stärkere Lemma: Jede Ecke von G ist Anfangspunkt einer Halbachse. Satz 1: Ist Geckentransitiv, dann liegt jede Ecke auf einer Achse. Satz 2: Ist A eine Achse in einem eckentransitiven Graphen G, dann hat G-A keine endliche Komponente. Sofort folgt Korollar: Seien A Achse in G und y Ecke von G- A. Dann liegt y auf einer mit A disjunkten Halbachse. Es wird untersucht, in welchen eckentransitiven Graphen G man im obig~n Korollar nHalbachse" durch "Achse" ersetzen kann. Falls G nicht 2-fach zusammenhängend ist, wird diese Frage vollständig beantwortet. Wir vermuten, daß in 2-fach zusamm~nhängenden Graphen mit unendlich vielen Enden eine durch y gehende mit A disjunkte Achse (für jedes A und jedes y in G- A) immer existiert.

R. J. WILSON (jointly with J. Rooney):. The mobility of a graph

In this expository talk we· discuss kinematic structures con­ structed from links and joints. We show how certain types of structure can be represented by direct or interchange graphs, and indicate how such graphs can be used to' enurnerate kinematic structures. After introducing the mobility of a

© - 20 -

structure (= # (degrees of freedom) - # (constraints) , we show how to define the mobility of a graph, and look in detail at two cases: (a) mobility 0 and binary links (Laman's theorem); (b) ~obility 1 and binary joints. Finally, we relate this material to the bracing of rectan- gular frameworks, using the result that such a framework is rigid if and only if an associated bipartite graph is connected.

Berichterstatter: H. Harborth, I. Mengersen •

© - 21 -

Tagungsteilnehmer

Prof. Dr. Andre Bouchet Dr. Ron Aharoni Dept. de Mathematiques Dept. of Mathematics Universite du Maing Technion F-72017 Le Mans, Cedex Hai f a Israel France

Dr. Brian Alspach Prof. Dr. Paul A. Catlln Simon Fraser University Dept. of Mathematics Dept. of Mathematics Wayne State Univ~rsit~

Bur n a b y, B.C., VSA 156 D e t r 0 i t, Mich. 48202' Canada U.5.A.

Dr. Thomas Andreae Guang-di Chen 11. Math. Institut Freie Universität Berlin University o~ Nanjing Königin-Luise-Str. 24-26 c/o Universität Hanqover Institut für Mathematik 1000 B e r 1 i n 33 Welfengarten 1 3000 H a n.n 0 ver

I. Ben-Arroyo Hartmann Prof. Dr. D~agos Cvetkqyic Math. Dept., Technion, Faculty of Electrical Technio'n City Engineering , University of Belgrade Hai f a Israel Bel g rad Hungary

Prof. Dr. Rainer Bodendiek Prof. Dr. Walter Deuber Pädagogische Hochschule' Kiel Fakultät, für -Mathematik:"­ Olshausenstr. 75 Universität Bielefeld Universitätsstr.. 1 - ", 2300 K i e 1 4800 B i eIe f eId '. 1

Prof. Dr. Adriah Bondy Prof ... Dr. Paul.. Erdös ',' Dept. Combinatorics and Math.Inst., of the Hungarian Optimization . Academy' of Sciences University of Waterloo Eötvös University Waterloo, Ontario N2L 3Gl H-1088 B u d a pes t Canada Hungary

© - 22 -

Prof. Dr. Andras Frank Prof. Dr. Andras Hajnal Research Inst. for Telecommuni­ Math.lnst. of the Hungarian cation Academy of Sciences Gabor A. u. 65 Eötvös University H-l026 B u d a pes t H-l088 B u d a pes t Hungary Hungary

Prof. Dr. Chris Godsil Prof. Dr. Rudolf Halin Dept. of Mathematics Mathematisches Seminar Simon Fraser University Universität Hamburg Bundess~r. 55 Bur n a b y, B.C. V5A 156 Canada 2000 H a m bur 9 13

Prof. Dr. Martin Grötschel Prof. Dr. Heiko Harborth Mathematisches Institut Bienroder Weg 47 Universität Augsburg 3300 Braunschweig Memminger 5tr. 6 8900 A u 9 s bur 9

Prof. Dr. A.J.W. Hilton Prof. Dr. Ivan Gutman Dept. of Mathematics Dept. of Mathematics Univ. of Reading ~niversity of Kragujevac R e a d i n g, RG6 2AX K rag u j eva c Great Britain Yugoslavia

Prof. Dr. Egmont Köhler Prof. Dr. Richard K. Guy Mathematisches Seminar University of Calgary Universität Hamburg Faculty of Science Bundesstr. 55 2500 University Dr. N.W. 2000 H a m bur 9 13 C a 1 gar y, Alb. Canada

Prof. Dr. Ervin Györi prof. Dr. Laszlo Lovasz Math.lnst. of the Hungarian c/o Institut für Operations Academy of Sciences Research Realtanoda u. 13-15 Universität Bonn Nassestr. 2 H-l053 B u d a pes t Hungary 5300 Bon n . 1

© - 23 -

Prof. Dr. Wolfgang Mader Prof. Dr. Janos Pach Institut für Mathematik Math.Inst. of the Hungarian Universität Hannover Academy of Sciences Welfengarten 1 . Realtanoda u. 13-15 3000 H a n n 0 ver H-1053 B u d a pes t Hungary

Prof. Dr. Torrence D. Parsons Dr. Ingrid Mengersen Ma thematics Dept. .. Werrastr. 17 Pennsylvania State University B h 9 3300 rau n s c w e i Uni ver 5 i t Y Park, PA 16802 U.S.A.

prof. Dr. Klaus-Peter Podewski" Prof. -Dr. Henry ~ey.niel Institut für Mathematik 43 rue Mazarine Universität Hannover F-75006 Par i s Welfengarten 1 France 3000 H a n n 0 ver

Prof. Dr. ReIf H. Möhring Prof. Dr. Gerhard Ringel Math.-Naturwiss. Fakultät Dept. of Mathematics' Techn. Hochschule Aachen . University of California San t a C r u z, Calif.95064 5100 A ach e n U.S.A.

_ Prof. Dr.· Henry Martyn Mulder Prof. Dr. Y. Roditty Vrye Universiteit School of Mathernatics Wisk.Sem. Tel Aviv University De Boelelaan Tel A v i v NL-1081 A m 5 t erd a m Netherlands Israel

·Prof. Dr. Jojtech Rödl p~ef.Dr. C.St.J.A. Nash-Williams FJFI - CVUT Dept. of Mathernatics katedra matematiky University of Reading Husova 5 R e a d i n g, RG6 2AX P rah a Great Britain CSSR

© - 24 -

Dr. Rüdiger Schmidt Prof. Dr. Klaus Wagner Wendum 30 Wodanstr. 57

2060 Bad 0 1 des 1 0 e 5000 K ö 1 n 91

Prof. Dr. Mark E. Watkins Prof. Dr. J. Schönheim School of Mathematics Dept. of Mathematics Tel Aviv university Syracuse University 200 Carneg:ie Tel A v i v S y r a c u s e, N.Y. 13210 Israel U.S.A.

Prof. Dr. A. Schrijver Prof. Dr. Richard Weiss Inst. voor Actuariaat en Tufts University Econometrie Univ. van Amsterdam M e d f 0 r d, Mass. 02155 Jodesbreestraat 23 U.S.A. NL-1011-NH A m s t erd a m

Dr. Heinz Schumacher Prof. Dr. Robin J. Wilson Pädagogische Ho~hschule Kiel 15 Chalfont Rd. Olshausenstr. 75 o x f 0 r d, OX2 6TL 2300 K i e 1 1 Great Britain

Prof. Dr. Karsten Steffens Prof. Dr. Joseph Zaks Institut für Mathematik Dept. of Mathematics Universität Hannover University of Haifa Welfengarten 1 Mount Carmel 3000 H a n n 0 ver Hai f a 31 999 Israel .

Prof. Dr. Bjarne Toft Matematisk Institut Odense Universitet DK-5230 0 den seM Danmark

©