Extremal problems for t-partite and t-colorable hypergraphs

Similar documents
On Control Problem Described by Infinite System of First-Order Differential Equations

arxiv: v1 [math.co] 4 Apr 2019

The shortest path between two truths in the real domain passes through the complex domain. J. Hadamard

Variance and Covariance Processes

336 ERIDANI kfk Lp = sup jf(y) ; f () jj j p p whee he supemum is aken ove all open balls = (a ) inr n, jj is he Lebesgue measue of in R n, () =(), f

7 Wave Equation in Higher Dimensions

Secure Frameproof Codes Through Biclique Covers

Math 301: The Erdős-Stone-Simonovitz Theorem and Extremal Numbers for Bipartite Graphs

ON 3-DIMENSIONAL CONTACT METRIC MANIFOLDS

@FMI c Kyung Moon Sa Co.

BMOA estimates and radial growth of B φ functions

Quantum Algorithms for Matrix Products over Semirings

Combinatorial Approach to M/M/1 Queues. Using Hypergeometric Functions

A note on characterization related to distributional properties of random translation, contraction and dilation of generalized order statistics

STUDY OF THE STRESS-STRENGTH RELIABILITY AMONG THE PARAMETERS OF GENERALIZED INVERSE WEIBULL DISTRIBUTION

SUFFICIENT CONDITIONS FOR MAXIMALLY EDGE-CONNECTED AND SUPER-EDGE-CONNECTED GRAPHS DEPENDING ON THE CLIQUE NUMBER

An intersection theorem for four sets

Lecture 18: Kinetics of Phase Growth in a Two-component System: general kinetics analysis based on the dilute-solution approximation

Sections 3.1 and 3.4 Exponential Functions (Growth and Decay)

Lecture 28: Convergence of Random Variables and Related Theorems

Deviation probability bounds for fractional martingales and related remarks

Journal of Inequalities in Pure and Applied Mathematics

Lecture 22 Electromagnetic Waves

Degree of Approximation of a Class of Function by (C, 1) (E, q) Means of Fourier Series

Order statistics and concentration of l r norms for log-concave vectors

Lecture-V Stochastic Processes and the Basic Term-Structure Equation 1 Stochastic Processes Any variable whose value changes over time in an uncertain

General Non-Arbitrage Model. I. Partial Differential Equation for Pricing A. Traded Underlying Security

Quasi-Randomness and the Distribution of Copies of a Fixed Graph

156 There are 9 books stacked on a shelf. The thickness of each book is either 1 inch or 2

THE MODULAR INEQUALITIES FOR A CLASS OF CONVOLUTION OPERATORS ON MONOTONE FUNCTIONS

MEEN 617 Handout #11 MODAL ANALYSIS OF MDOF Systems with VISCOUS DAMPING

The sudden release of a large amount of energy E into a background fluid of density

, on the power of the transmitter P t fed to it, and on the distance R between the antenna and the observation point as. r r t

Reichenbach and f-generated implications in fuzzy database relations

Fuzzy Hv-submodules in Γ-Hv-modules Arvind Kumar Sinha 1, Manoj Kumar Dewangan 2 Department of Mathematics NIT Raipur, Chhattisgarh, India

Monochromatic Infinite Sumsets

Research Article A Note on Multiplication and Composition Operators in Lorentz Spaces

Representing Knowledge. CS 188: Artificial Intelligence Fall Properties of BNs. Independence? Reachability (the Bayes Ball) Example

4 Sequences of measurable functions

The Central Limit Theorems for Sums of Powers of Function of Independent Random Variables

An Automatic Door Sensor Using Image Processing

K. G. Malyutin, T. I. Malyutina, I. I. Kozlova ON SUBHARMONIC FUNCTIONS IN THE HALF-PLANE OF INFINITE ORDER WITH RADIALLY DISTRIBUTED MEASURE

ENGI 4430 Advanced Calculus for Engineering Faculty of Engineering and Applied Science Problem Set 9 Solutions [Theorems of Gauss and Stokes]

Some Ramsey results for the n-cube

Computer Propagation Analysis Tools

Additive Approximation for Edge-Deletion Problems

On The Estimation of Two Missing Values in Randomized Complete Block Designs

New problems in universal algebraic geometry illustrated by boolean equations

Chapter 3: Theory of Modular Arithmetic 38

MATHEMATICAL FOUNDATIONS FOR APPROXIMATING PARTICLE BEHAVIOUR AT RADIUS OF THE PLANCK LENGTH

LOGARITHMIC ORDER AND TYPE OF INDETERMINATE MOMENT PROBLEMS

arxiv: v1 [math.co] 4 May 2017

Upper Bounds for Tura n Numbers. Alexander Sidorenko

The Chromatic Villainy of Complete Multipartite Graphs

On the ratio of maximum and minimum degree in maximal intersecting families

POSITIVE SOLUTIONS WITH SPECIFIC ASYMPTOTIC BEHAVIOR FOR A POLYHARMONIC PROBLEM ON R n. Abdelwaheb Dhifli

Method for Approximating Irrational Numbers

Monochromatic Wave over One and Two Bars

arxiv: v1 [math.nt] 28 Oct 2017

Reinforcement learning

Vehicle Arrival Models : Headway

We just finished the Erdős-Stone Theorem, and ex(n, F ) (1 1/(χ(F ) 1)) ( n

EXERCISES FOR SECTION 1.5

arxiv: v1 [math.co] 6 Mar 2008

Wide-sense fingerprinting codes and honeycomb arrays

Brief summary of functional analysis APPM 5440 Fall 2014 Applied Analysis

Oscillation of an Euler Cauchy Dynamic Equation S. Huff, G. Olumolode, N. Pennington, and A. Peterson

A Bijective Approach to the Permutational Power of a Priority Queue

EVENT HORIZONS IN COSMOLOGY

INDEPENDENT SETS IN GRAPHS WITH GIVEN MINIMUM DEGREE

New lower bounds for the independence number of sparse graphs and hypergraphs

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality

Solution to HW 3, Ma 1a Fall 2016

Math 115 Final Exam December 14, 2017

11!Hí MATHEMATICS : ERDŐS AND ULAM PROC. N. A. S. of decomposiion, properly speaking) conradics he possibiliy of defining a counably addiive real-valu

arxiv: v2 [math.pr] 19 Feb 2016

ON INDEPENDENT SETS IN PURELY ATOMIC PROBABILITY SPACES WITH GEOMETRIC DISTRIBUTION. 1. Introduction. 1 r r. r k for every set E A, E \ {0},

1 Solutions to selected problems

Lecture 17: Kinetics of Phase Growth in a Two-component System:

arxiv: v1 [math.ca] 15 Jan 2019

Probablistically Checkable Proofs

AVERAGING OF ENTIRE FUNCTIONS OF IMPROVED REGULAR GROWTH WITH ZEROS ON A FINITE SYSTEM OF RAYS R.V. Khats

Lecture 20: Riccati Equations and Least Squares Feedback Control

A STOCHASTIC MODELING FOR THE UNSTABLE FINANCIAL MARKETS

arxiv:math/ v1 [math.nt] 3 Nov 2005

On the ratio of maximum and minimum degree in maximal intersecting families

Deterministic vs Non-deterministic Graph Property Testing

Final Spring 2007

arxiv: v1 [math.ca] 25 Sep 2013

Turán Numbers of Vertex-disjoint Cliques in r- Partite Graphs

ON THE INVERSE SIGNED TOTAL DOMINATION NUMBER IN GRAPHS. D.A. Mojdeh and B. Samadi

Today - Lecture 13. Today s lecture continue with rotations, torque, Note that chapters 11, 12, 13 all involve rotations

PHYS PRACTICE EXAM 2

A THREE CRITICAL POINTS THEOREM AND ITS APPLICATIONS TO THE ORDINARY DIRICHLET PROBLEM

Numerical Integration

MATH 5720: Gradient Methods Hung Phan, UMass Lowell October 4, 2018

Dual Hierarchies of a Multi-Component Camassa Holm System

Central Coverage Bayes Prediction Intervals for the Generalized Pareto Distribution

arxiv: v2 [math.ag] 4 Jul 2012

Finish reading Chapter 2 of Spivak, rereading earlier sections as necessary. handout and fill in some missing details!

Transcription:

Exemal poblems fo -paie and -coloable hypegaphs Dhuv Mubayi John Talbo June, 007 Absac Fix ineges and an -unifom hypegaph F. We pove ha he maximum numbe of edges in a -paie -unifom hypegaph on n veices ha conains no copy ( of F is c n,f ) + o(n ), whee c,f can be deemined by a finie compuaion. We explicily define a sequence F 1, F,... of -unifom hypegaphs, and pove ha he maximum numbe of edges in a -chomaic -unifom hypegaph on n veices ( conaining no copy of F i is α n,,i ) + o(n ), whee α,,i can be deemined by a finie compuaion fo each i 1. In seveal cases, α,,i is iaional. The main ool used in he poofs is he Lagangian of a hypegaph. 1 Inoducion An -unifom hypegaph o -gaph is a pai G = (V, E) of veices, V, and edges E ( V ), in paicula a -gaph is a gaph. We denoe an edge {v 1, v,..., v } by v 1 v v. Given -gaphs F and G we say ha G is F -fee if G does no conain a copy of F. The maximum numbe of edges in an F -fee -gaph of ode n is ex(n, F ). Fo = and F = K s (s 3) Depamen of Mahemaics, Saisics and Compue Science, Univesiy of Illinois, Chicago, IL 60607, and Depamen of Mahemaical Sciences, Canegie-Mellon Univesiy, Pisbugh, PA 1513. Email: mubayi@mah.uic.edu. Reseach suppoed in pa by NSF gans DMS-040081 and 0653946 and an Alfed P. Sloan Reseach Fellowship. Depamen of Mahemaics, UCL, London, WC1E 6BT, UK. Email: albo@mah.ucl.ac.uk. This auho is a Royal Sociey Univesiy Reseach Fellow. 1

his numbe was deemined by Tuán [T41] (ealie Manel [M07] found ex(n, K 3 )). Howeve in geneal (even fo = ) he poblem of deemining he exac value of ex(n, F ) is beyond cuen mehods. The coesponding asympoic poblem is o deemine he Tuán densiy ex(n,f ) of F, defined by π(f ) = lim n (his always exiss by a simple aveaging agumen ( n ) due Kaona e al. [KNS64]). Fo -gaphs he Tuán densiy is deemined by he chomaic numbe of he fobidden subgaph F. The explici elaionship is given by he following fundamenal esul. Theoem 1 (Edős Sone Simonovis [ES46],[ES66]). If F is a -gaph hen π(f ) = 1 1. χ(f ) 1 When 3, deemining he Tuán densiy is difficul, and hee ae only a few exac esuls. Hee we conside some closely elaed hypegaph exemal poblems. Call a hypegaph H -paie if is veex se can be paiioned ino classes, such ha evey edge has a mos one veex in each class. Call H -coloable, if is veex se can be paiioned ino classes so ha no edge is eniely conained wihin a class. Definiion. Fix, and an -gaph F. Le ex (n, F ) (ex (n, F )) denoe he maximum numbe of edges in a -paie (-coloable) -gaph on n veices ha conains no copy of F. The -paie Tuán densiy of F is π (F ) = lim n ex (n, F )/ ( n ) and he -chomaic Tuán densiy of F is π (F ) = lim n ex (n, F )/ ( n ). Noe ha i is easy o show ha hese limis exis. In his pape, we deemine π (F ) fo all -gaphs F and deemine π (F ) fo an infinie family of -gaphs (peviously no nonivial value of π (F ) was known). In many cases ou examples yield iaional values of π (F ). Fo he usual Tuán densiy, π(f ) has no been poved o be iaional fo any F, alhough hee ae seveal conjecues saing iaional values. In ode o descibe ou esuls, we need he concep of G-colouings which we inoduce now. If F and G ae hypegaphs (no necessaily unifom) hen F is G-colouable if hee exiss c : V (F ) V (G) such ha c(e) E(G) wheneve e E(F ). In ohe wods, F is G-colouable if hee is a homomophism fom F o G. Le K () K () denoe he complee -gaph of ode. Then an -gaph F is -paie if F is -colouable, and F is -colouable if i is H () -colouable whee H () is he (in geneal non-unifom) hypegaph consising of all subses A {1,,..., } saisfying A ). The chomaic numbe of F is χ(f ) = min{ 1 : F is -colouable}. Noe ha while a

-gaph is -colouable iff i is -paie his is no longe ue fo 3, fo example K (3) 4 is -colouable bu no -paie o 3-paie. Le G () denoe he collecion of all -veex -gaphs. A ool which has poved vey useful in exemal gaph heoy and which we will use lae is he Lagangian of an -gaph. Le S = { x R : x i = 1, x i 0 fo 1 i }. i=1 If G G () and x S hen we define λ(g, x) = x v1 x v x v. v 1 v v E(G) The Lagangian of G is max x S λ(g, x). The fis applicaion of he Lagangian o exemal gaph heoy was due o Mozkin and Sauss who gave a new poof of Tuán s heoem. We ae now eady o sae ou main esul. Theoem 3. If F is an -gaph and hen π (F ) = max{!λ(g) : G G () and F is no G-colouable}. As an example of Theoem 3, suppose ha = 4, = 3, and F = K (3) 4. Le H denoe he unique 3-gaph wih fou veices and hee edges. Now F is F -coloable, bu i is no H-coloable, and he lagangian λ(h) of H is 4/81, achieved by assigning he degee hee veex a weigh of 1/3 and he ohe hee veices a weigh of /9. Consequenly, Theoem 3 says ha he maximum numbe of edges in an n-veex 4-paie 3-gaph conaining no copy of K (3) 4 is (8/7) ( n 3) + o(n 3 ). This is clealy achievable, by he 4-paie 3-gaph wih pa sizes n/3, n/9, n/9, n/9, wih all possible iples beween hee pas ha include he lages (of size n/3), and no iples beween he hee small pas. Chomaic Tuán densiies wee peviously consideed in [T07] whee hey wee used o give an impoved uppe bound on π(h), whee H is defined in he pevious paagaph. Howeve no non-ivial chomaic Tuán densiies have peviously been deemined. each we ae able o give an infinie sequence of -gaphs whose -chomaic Tuán densiies ae deemined exacly. Fo l and define β,,l := max{λ(g) : G is a -coloable -gaph on l veices}. 3 Fo

I seems obvious ha β,,l is achieved by he -chomaic -gaph of ode l wih all colo classes of size l/ o l/ and all edges pesen excep hose wihin he classes. Noe ha if l hen his would give β,,l = (( ) l ( l/ )) 1 l. Howeve, we ae only able o pove his fo =, 3. If he above saemen is ue, hen β,,l can be compued by calculaing he maximum of an explici polynomial in one vaiable ove he uni ineval. In any case i can be obained by a finie compuaion (fo fixed,, l). Le α,,l =!β,,l. Theoem 4. Fix l. Le L () be he -gaph obained fom he complee gaph K by enlaging each edge wih a se of new veices. If hen whee α,,l is defined above. π (L () ) = α,,l The emainde of he pape is aanged as follows. In he nex secion we pove Theoem 3 and in he las secion we pove Theoem 4 and he saemens abou compuing β,,l, fo =, 3. Poof of Theoem 3 If G G () and x = (x 1,..., x ) Z + hen he x-blow-up of G is he -gaph G( x) consuced fom G by eplacing each veex v by a class of veices of size x v and aking all edges beween any classes coesponding o an edge of G. Moe pecisely we have V (G( x)) = X 1 X, X i = x i and E(G( x)) = {{v i1 v i v i } : v ij X ij, {i 1 i i } E(G)}. If x = (s, s,..., s) and G = K () hen G( x) is he complee -paie -gaph wih class size s, denoed by K () (s). Noe ha if F and G ae boh -gaphs hen F is G-colouable iff hee exiss x Z + such ha F G( x). An -gaph G is said o be coveing if each pai of veices in V (G) is conained in a common edge. If W V and G is an -gaph wih veex V hen G[W ] is he induced subgaph of G fomed by deleing all veices no in W and emoving all edges conaining hese veices. 4

Lemma 5 (Fankl and Rödl [FR84]). If G is an -gaph of ode n hen hee exiss y S n wih λ(g) = λ(g, y), such ha if P = {v V (G) : y v > 0} hen G[P ] is coveing. Supesauaion fo odinay Tuán densiies was shown by Edős [E71]. The poof fo G- chomaic Tuán densiies is essenially idenical bu fo compleeness we give i. We equie he following classical esul. Theoem 6 (Edős [E64]). If and 1 hen ex(n, K () ()) = O(n λ, ), wih λ, > 0. Lemma 7 (Supesauaion). Fix. If G is an -gaph, H is a finie family of - gaphs, s 1 and s = (s, s,..., s) hen π (H( s)) = π (H) (whee H( s) = {H( s) : H H}). Poof: Le p = max{ V (H) : H H}. By adding isolaed veices if necessay we may suppose ha evey H H has exacly p veices. Fis we claim ha if F is an n-veex -gaph wih densiy a leas α + ɛ, whee α, ɛ > 0, and m n hen a leas ɛ ( n m) of he m-veex induced subgaphs of F have densiy a leas α + ɛ. To see his noe ha if i fails o hold hen ) ( ) n ( )( ) ( ) ( n m n m ( n m (α + ɛ) which is impossible. W ( V (F ) m ) e(f [W ]) < ɛ m + (1 ɛ) m (α + ɛ) Le ɛ > 0 and suppose ha F is an n-veex -gaph wih densiy a leas π (H) + ɛ. We need o show ha if n sufficienly lage hen F conains a copy of H( s). Le m m(ɛ) be sufficienly lage ha any -paie m-veex -gaph wih densiy a leas π (H)+ɛ conains a copy of some H H. We say ha W ( ) V (F ) m is good if F [W ] conains a copy of some H H. By he claim a leas ɛ ( ( n m) m-ses ae good, so if δ = ɛ/ H hen a leas δ n m) m-ses conain a fixed H H. Thus he numbe of p-ses U V (F ) such ha F [U] H is a leas δ ( ) n m ) = δ( ) n p ( m ). (1) p ( n p m p Le J be he p-gaph wih veex se V (F ) and edge se consising of hose p-ses U V (F ) such ha F [U] H. Now, by Theoem 6, ex (n, K () ()) ex(n, K () ()) = O(n λ, ), whee λ, > 0. Hence (1) implies ha fo any if n is sufficienly lage hen K p (p) () J. ), 5

Finally conside a colouing of he edges of K p (p) () wih p! diffeen colous, whee he colou of he edge is given by he ode in which he veices of H ae embedded in i. By Ramsey s heoem if is sufficienly lage hen hee is a copy of K p (p) (s) wih all edges he same colou. This yields a copy of H ( s) in F as equied. Poof of Theoem 3. Le α, = max{!λ(g) : G G () is well-defined since G () ( ) is finie.) and F is no G-colouable}. (This If G G () and F is no G-colouable hen fo any x Z + we have F G( x). Le y S saisfy λ(g, y) = λ(g). Fo n 1 le x n = ( y 1 n,..., y n ) Z +. If G n = G( x n ) hen e(g n ) lim ( n n =!λ(g). ) Moeove since each G n is F -fee, -paie and of ode a mos n we have π (F )!λ(g). Hence π (F ) α,. Le H(F ) = {H G () : F is H-colouable}. I is sufficien o show ha π (H(F )) α,. () Indeed, if we assume ha () holds, hen le s 1 be minimal such ha evey H H(F ) saisfies F H( s), whee s = (s, s,..., s). (Noe ha s exiss since F is H-colouable fo evey H H(F )). Now by supesauaion (Lemma 7) if ɛ > 0, hen any -paie -gaph G n wih n n 0 (s, ɛ) veices and densiy a leas α, + ɛ will conain a copy of H( s) fo some H H(F ). In paicula G n conains F and so π (F ) α,. Le π (H(F )) = γ and ɛ > 0. If n is sufficienly lage hee exiss an H(F )-fee, -paie -gaph G n of ode n saisfying!e(g n ) γ ɛ. n Taking y = (1/n, 1/n,..., 1/n) S n we have!λ(g n )!λ(g n, y) =!e(g n) n γ ɛ. Now Lemma 5 implies ha hee exiss z S n saisfying λ(g n ) = λ(g n, z) and G n [P ] is coveing whee P = {v V (G) : z v > 0}. 6

Since G n is -paie, we conclude ha G n [P ] has a mos veices. Moeove, G n is H(F )- fee and so G n [P ] H(F ). Thus F is no G n [P ]-coloable, and we have γ ɛ!λ(g n [P ]) α,. Thus π (H(F )) α, + ɛ fo all ɛ > 0. Hence () holds and he poof is complee. 3 Infiniely many chomaic Tuán densiies Fo l, le K () l be he family of -gaphs wih a mos ( l ) edges ha conain a se S, called he coe, of l veices, wih each pai of veices fom S conained in an edge. Noe ha L () K(). We need he following Lemma ha was poved in [M06]. Fo compleeness, we epea he poof below. Lemma 8. If K K (), s = ( ) () + 1 and s = (s, s,..., s) hen L K( s). Poof. We fis show ha L () L(( ) () + 1) fo evey L K. Pick L K(), and le L = L( ( ) + 1). Fo each veex v V (L), suppose ha he clones of v ae v = v 1, v,..., v ( )+1. In paicula, idenify he fis clone of v wih v. Le S = {w 1,..., w } V (L) be he coe of L. Fo evey 1 i < j l + 1, le E ij L wih E ij {w i, w j }. Replace each veex z of E ij {w i, w j } by z q whee q > 1, o obain an edge E ij L. Coninue his pocedue fo evey i, j, making sue ha wheneve we encoune a new edge i inesecs he peviously encouneed edges only in L. Since he numbe of clones is ( ) + 1, his pocedue can be caied ou successfully and esuls in a copy of L () wih coe S. Theefoe L() L = L( ( ) + 1). Consequenly, Lemma 7 implies ha π(l () ) π(k() ). Poof of Theoem 4. Le l and. We will pove ha π (K () ) = α,,l. (3) The heoem will hen follow immediaely fom Lemmas 7 and 8. Le B,,l = {G : G is a -colouable K () -fee -gaph}. Claim. max{λ(g) : G B,,l } = β,,l = α,,l /!. Poof of Claim. If G B,,l has ode n hen Lemma 5 implies ha hee is y S n such ha λ(g) = λ(g, y) wih G[P ] coveing, whee P = {v V (G) : y v > 0}. Since G is 7

K () -fee, we conclude ha P = p l. Hence hee is H B,,l such ha λ(h) = λ(g) and H has ode a mos l. Consequenly, max{λ(g) : G B,,l } β,,l. Fo he ohe inequaliy, we jus obseve ha an l veex -gaph mus be K () -fee. Now we can quickly complee he poof of he heoem by poving (3). Fo he uppe bound, obseve ha if G B,,l has ode n hen by he Claim e(g) n λ(g) α,,l! and so π (K () ) α,,l. Fo he lowe bound, suppose ha G B,,l has ode p and saisfies λ(g) = β,,l. Then hee exiss y S p such ha λ(g, y) = λ(g) = β,,l. Fo n p define y n = ( y 1 n,..., y p n ). Now {G( y n )} n=p is a sequence of -colouable K () -fee -gaphs and hence π (K () ) lim e(g n ) ( n n =!λ(g) = α,,l. ) Now we pove ha β,,l can be compued by only consideing maximum -coloable -gaphs wih almos equal pa sizes when =, 3. The case = follows ivially fom Lemma 5 so we conside he case = 3. Theoem 9. Fix l. Then β 3,,l is achieved by he -chomaic 3-gaph of ode l wih all colo classes of size l/ o l/ and all edges pesen excep hose wihin he classes. Remak: Noe ha if l hen his implies ha β 3,,l = ( ( ) ( l 3 l/ ) 3 ) 1 Poof. Le G be a -chomaic 3-gaph of ode l. We may suppose (by adding edges as equied) ha V (G) = V 1 V V and ha all edges no conained in any V i ae pesen. We may also suppose ha V 1 V V. Le x S p saisfy λ(g, x) = λ(g). If v, w V i and x v > x w hen fo a suiable choice of δ > 0 we can incease λ(g, x) by inceasing x w by δ and deceasing x v by δ. Hence we may suppose ha hee ae x 1,..., x such ha all veices in V i eceive weigh x i. Le l = b + c, 0 c <. To complee he poof we need o show ha all of he V i have ode b o b + 1. Suppose, fo a conadicion, ha hee exis V i and V j wih a i = V i, a j = V j and a i a j +. Moving a veex v fom V i o V j and inseing all new allowable edges (i.e. hose which conain v and veices fom V i \{v}) while deleing any edges which l 3. 8

now lie in V j we canno incease λ(g, x). This implies ha ( ) ( ) aj x i x ai 1 j x 3 i, (4) and so in paicula x i < x j. Le G denoe his new -colouable 3-gaph. We give a new weighing y fo G by seing a i x i /(a i 1), v V i, y v = a j x j /(a j + 1), v V j, x k, v V k and k i, j. I is easy o check ha y S l is a legal weighing fo G. We will deive a conadicion by showing ha λ( G) λ( G, y) > λ(g, x) = λ(g). If w = a i x i + a j x j = (a i 1)y i + (a j + 1)y j hen (( ) ( ) λ( G, ai 1 y) λ(g, x) = (1 w) yi aj + 1 + yj + (a i 1)(a j + 1)y i y j ( ) ( ) ( ) ai x aj i )x j ai 1 a i a j x i x j + (a j + 1)yi y j + ( ) ( ) ( ) aj + 1 (a i 1)y i yj ai a j x aj i x j a i x i x j = (1 w) ( aj x j a j + 1 a ix i a i 1 Using (4) i is easy o check ha his is sicly posiive. ) + a ia j x i x j ( xj a j + 1 Coollay 10. The -chomaic Tuán densiy can ake iaional values. x ) i. a i 1 Poof. We conside β 3,,k fo k 3. In fac, we focus of β 3,,6, he maximum densiy of a -chomaic 3-gaph ha conains no copy of K (3) 6. By he pevious Theoem, his is 6 imes he lagangian of he 3-gaph wih veex se {a, a, a, b, b } and all edges pesen excep {a, a, a }. Assigning weigh x o he a s and weigh y o he b s, we mus maximize 6(6x y + 3xy ) subjec o 3x + y = 1 and 0 x 1/3. A sho calculaion shows ha he choice of x ha maximizes his expession is ( 13 )/9, and his esuls in an iaional value fo he lagangian. Simila compuaions hold fo lage k as well. 9

Refeences [E64] P. Edős, On exemal poblems of gaphs and genealized gaphs, Isael J. Mah. (1964), 183 190. [E71] P. Edős, On some exemal poblems on -gaphs, Disc. Mah. 1 (1971), 1 6. [ES66] [ES46] P. Edős and M. Simonovis, A limi heoem in gaph heoy, Sudia Sci. Ma. Hung. Acad. 1 (1966), 51 57. P. Edős and A.H. Sone, On he sucue of linea gaphs, Bull. Ame. Mah. Soc. 5(1946), 1087 1091. [FR84] P. Fankl and V. Rödl, Hypegaphs do no jump, Combinaoica 4 (1984), 149 159. [KNS64] G. Kaona, T. Nemez and M. Simonovis, On a poblem of Tuan in he heoy of gaphs (in Hungaian) Ma. Lapok 15 (1964), 8 38. [M07] W. Manel, Poblem 8, Wiskundige Opgaven, 10 (1907), 60-61. [MS65] [M06] [T07] T. Mozkin and E. Sauss. Maxima fo gaphs and a new poof of a heoem of Tuan, Canadian Jounal of Mahemaics, 17:533 540, 1965. D. Mubayi, A hypegaph exension of Tuán s heoem, J. Combin. Theoy, Se. B 96 (006) 1 134. J. Talbo, Chomaic Tuán poblems and a new uppe bound fo he Tuán densiy of K4. Euop. J. Comb. (o appea) (007?) [T41] P. Tuán, On an exemal poblem in gaph heoy, Ma. Fiz. Lapok 48 (1941) 10