arXiv:1903.02099v2 [math.CO] 18 Dec 2020 1.2. nietdgahof graph undirected rp disa cci retto,i sntpsil for possible not is it orientation, acyclic an admits graph ob an be to { an G graph yls nohrwrs h retdgraph oriented the words, other In cycles. ( h aeo nietdgah,a lmn of element an graphs, undirected of case the yRme [ Ramsey by where otisamncrmtccp of copy monochromatic a contains where n #2018/05557-7. and 1.1. ori for notion this definitions. of few variant a a with for problems same graph the random study We a that property the for function ncmiaois(e ..tednmcsre fRadziszows of survey dynamic the e.g. (see combinatorics in ,v u, ,v u, Teato cnwegsspotfo ã al eerhFo Research Paulo São from support acknowledges author *The Date ie graphs Given → arc ) ∈ } retdRme number. Ramsey Oriented irpsadoine graphs. oriented and Digraphs H ∈ ~ of V 001/1 2:22am 2020/12/21, : G ASYTP RBESFROINAIN FGRAPHS OF ORIENTATIONS FOR PROBLEMS RAMSEY-TYPE odffrfo h nietdcs.An case. undirected the from differ to Abstract. retto of orientation htteeeit graph a exists there that asynme fagraph a of number Ramsey rpsadaatn nagmn fHn etr öladScha and Rödl Retter, Hàn, hyper of the argument applying an by adapting cycle, for and the bound graphs of upper orientation an acyclic prove an we of and model, isometric the consider graph hehl ucinfrteevent the for function threshold E E odnt hteeyoinaino h de of edges the of orientation every that denote to G sasto etcsand vertices of set a is orientation = ~ ′ implies utemr,when Furthermore, . fadol if only and if G R 23 H ~ n ( n rő n zkrs[ Szekeres and Erdős and ] H eoe that denotes with G =inf := ) ( ie naylcoine graph oriented acyclic an Given and ,u v, G G ~ h fagraph a of a noine oyof copy oriented an has ewl lasdnt irp yacptllte with letter capital a by digraph a denote always will We . ) etcs where vertices, H ∈ / { ewrite we , ( n ,v u, E ∈ o every for G ) G N RN AQAOT CAVALAR* PASQUALOTTO BRUNO ∈ H sagahon graph a is E G : forder of G eso that show we , ~ E hr xssagraph a exists there H sastsc that such set a is ( = sa retdgah ewrite we graph, oriented an is { G or H ie graph a Given G h asynumber Ramsey The . 1. ( ,v u, V → ( ,p n, sisudryn nietdgah eas td the study also We graph. undirected underlying its is ,u v, n ′ E , Introduction H A 12 ) satisfying ∈ H H ~ ~ ) ′ → odnt hteeytoclrn fteegsof edges the of two-coloring every that denote to .Fnigbud for bounds Finding ]. ietdgraph directed ) V edefine We . E ∈ utb cci.Oemyas s o onson bounds for ask also may One acyclic. be must if n H G oevr noine graph oriented an Moreover, . ~ R 1 ~ E H scle an called is ~ retdgraph oriented V } ( ( etcs hsnme a rvdt efinite be to proved was number This vertices. nti ae esythat say we case, this In . H ,p n, ~ n graph a and ntebnma admgah ial,we Finally, graph. random binomial the in = ) G 6 G ) E V → G satisfies 2 ′ n noine graph oriented an and R ⊆ R ~ H G n,frevery for and, ~ = ( ( nain(AEP,gat #2015/26678-9 grants (FAPESP), undation H G ( H eoigby Denoting . ~ rp otie em nrandom in lemma container graph V → G edge or ) ) c R otisacp of copy a contains n × log stesals number smallest the as G i[ ki H ( h smti asynumber Ramsey isometric the nain fgah.Ltu begin us Let graphs. of entations ~ digraph G H uhthat such cht. ewrite we , 2 V G oee,i a lob called be also may it however, ; ~ h ( oocrif occur to ) 22 ) ,p n, o vr cci oriented acyclic every for ( = { \ fagraph a of R G ) oevr h threshold the Moreover, ]). ( ) ( H ,E V, odnt h underlying the denote to ,v v, → G ~ ,v u, R ) G G ( H sacasclproblem classical a is sapair a is : ) H ) H → → ) ∈ G ~ sadgahwhere digraph a is swl-tde [ well-studied is v h classical the G ~ H V otisdirected contains H sthe is H ∈ ~ H H ( = ~ ~ } sdfie as defined is V =
fevery if e swrite us let , ic every Since . n , } . ,E V, V G uta in as Just . ~ such underlying ′ ehave we , ( = ) ssaid is ,E V, 24 G ]. ) the oriented Ramsey number R~(H~ ), which is defined as

R~ (H~ ) := inf n N : there exists a graph G = Gn such that G H~ . ∈ → n o We can define R~ (H~ ) equivalently as the smallest natural number n such that every tournament on n vertices contains H~ . To the best of our knowledge, this number was first studied by Erdős and Moser, who proved the following theorem.

Theorem 1.1 (Erdős and Moser [11]). Let K~ k be the transitive tournament on k vertices. We have (k−1)/2 k−1 2 6 R~ (K~ k) 6 2 .

Since every acyclic oriented graph H~ is contained in the transitive tournament with the same number of vertices, this implies that R~ (H~ ) is finite. We now briefly survey a few bounds for the oriented Ramsey number of orientations of paths, cycles and trees. Let H~ be the directed path on k vertices. A well-known result obtained independently by Gallai, Hasse, Roy and Vitaver (see, for example, Theorem 14.5 of Bondy and Murty [5]) implies that R~ (H~ )= k. Now suppose that H~ is any orientation of the undirected path with k > 2128 vertices. The result above was improved by Thomason [28], who showed that we also have R~ (H~ )= k in this case. This condition was later weakened to k > 9 by Havet and Thomassé [15]. Suppose now that H~ is an acyclic orientation of the cycle on k vertices. The same work due to Thomason [28] also showed that, if k > 2128, then R~ (H~ )= k. Havet and Thomassé’s work [15] again reduced this requirement to k > 68. Moreover, Heydemann, Sotteau and Thomassen [16] proved that, if a graph G has at least (k 1)(k 2) + 3 edges, then G H~ . One can therefore − − → do a simple calculation to show that R~(H~ ) 6 √2k, thus proving that a linear bound for the oriented Ramsey number also holds for cycles of size smaller than 68. Suppose that H~ is an oriented tree on k vertices. A conjecture due to Sumner (see [13]) says that R~ (H~ ) 6 2k 2. This inequality is optimal, as there are oriented trees for which their − oriented Ramsey number is at least 2k 2. A linear upper bound to R~ (H~ ) was first proved − by Häggkvist and Thomason [13], who showed that R~ (H~ ) 6 12k. El Sahili [10] proved that R~ (H~ ) 6 3k 3 for all oriented trees on k vertices, though its known that R~ (H~ )= k for almost − all such trees, as proved by Mycroft and Naia [19]. El Sahili’s bound is currently the best upper bound when k is small, but, for large enough k, Sumner’s conjecture was proved by Kühn, Mycroft and Osthus [17]. A good survey about the existing work on bounds for the oriented Ramsey number of trees can be found in the Introduction of [20]. Finally, we remark that a work of Bloom and Burr [4] connects the oriented Ramsey number of an oriented graph with its family of homomorphisms. We also remark that, in most of the literature, the oriented Ramsey number is not studied with the language of Ramsey theory, but rather it appears under the guise of “unavoidable digraphs” and related terminology. We believe that connecting this notion to the well-studied Ramsey theory can provide many benefits. Unlike the classical Ramsey number, about which much is known, little else has been published on bounds for the oriented Ramsey number. In Section 2, we apply results and concepts from Conlon, Fox, Lee, and Sudakov [9] and Balko, Cibulka, Král and Kynčl [1], so as to show that 2 R~ (H~ ) 6 2R(H)c log h. This gives an automatic upper bound on the oriented Ramsey number of every oriented graph by considering the Ramsey number of its underlying undirected graph,

2 with only logarithmic loss in the exponent. We leave open the question of whether there exists an oriented graph for which this bound is tight.

1.3. An oriented Ramsey theorem for random graphs. For a graph H, we denote by m2(H) its 2-density, defined as e(F ) 1 m2(H) := max − . F ⊆H,v(F )>3 v(F ) 2 − Consider also the binomial random graph G(n,p), which is the random graph of order n in which each edge appears independently with probability p. A celebrated result of Rödl and Ruciński [24] determined, for an undirected graph H, the threshold function for G(n,p) H. → (Here we state only the 1-statement.)

Theorem 1.2 (Rödl and Ruciński [24]). Let H be a graph. There exists a constant C = C(H) such that, if p > Cn−1/m2(H), then

lim P[G(n,p) H] = 1. n→∞ →

Define m2(H~ ) := m2(H). In Section 4, we prove the following version of Theorem 1.2 for acyclic oriented graphs.

Theorem. Let H~ be an acyclic oriented graph. There exists a constant C = C(H~ ) such that, if ~ p > Cn−1/m2(H), then lim P G(n,p) H~ = 1. n→∞ → h i Adapting arguments from Nenadov and Steger [21], our proof of Theorem 4.1 makes use of the hypergraph container lemma of Balogh, Morris and Samotij [2] and Saxton and Thomason [27]. In Section 3, we develop the necessary container theory for digraphs that allows us to prove Theorem 4.1 in Section 4. The technique of using hypergraph containers in random graphs for Ramsey problems has recently been employed by Hàn, Retter, Rödl and Schacht [14], Rödl, Ruciński and Schacht [25] and Conlon, Dellamonica, La Fleur, Rödl and Schacht [8]. Our approach is also inspired by theirs, and some resemblance to their arguments is to be expected.

1.4. Isometric oriented Ramsey number. Finally, we consider the isometric oriented Ram- sey number R~ iso(H~ ) of an acyclic oriented graph H~ , a concept first introduced by Banakh, Idzik, Pikhurko, Protasov and Pszczoła [3]. For an undirected graph G, we denote by d (u, v) the distance between two vertices u, v G ∈ V (G). Given two oriented graphs H~ and F~ , we say that a copy f : V (H~ ) V (F~ ) of H~ in F~ is → an isometric copy if d (x,y)= d (f(x),f(y)) for every x,y V (H~ ). Note that the distance is H F ∈ taken with respect to the underlying undirected graphs. iso Given an oriented graph H~ and a graph G, we write G H~ if every orientation of G has an −→ isometric oriented copy of H~ . The isometric oriented Ramsey number R~ iso(H~ ) is defined as iso R~ (H~ ) := inf n N : there exists a graph G = Gn such that G H~ . iso ∈ −→ n o It was proved in [3, Theorem 2.1] that the isometric oriented Ramsey number of acyclic oriented graphs is always finite. In Section 5, we devise a bound for R~ iso(H~ ) when H~ is an acyclic

3 orientation of a cycle, adapting a construction of Hàn, Retter, Rödl, and Schacht [14]. In particular, we prove the following theorem.

Theorem. There exists a positive constant c such that the following holds. Let H~ be an acyclic orientation of Ck and set R := R~ (H~ ). Then

12k3 8k2 R~ iso(H~ ) 6 ck R . (1)

The proof also makes use of the container results we will develop in Section 3.

2. Bounds for the oriented Ramsey number

Before stating our bounds, we introduce the concept of ordered graphs and ordered Ramsey numbers, recently studied in Balko, Cibulka, Král and Kynčl [1] and Conlon, Fox, Lee, and Sudakov [9]. ′ ′ An ordered graph G is a pair G = (G ,

Theorem 2.1 (Conlon, Fox, Lee, and Sudakov [9]). There exists a constant c such that, for every ordered graph H on n vertices, we have

c log2 n R<(H) 6 R(H) .

More precise bounds for R<(H) for specific classes of ordered graphs can be found in Conlon, Fox, Lee, and Sudakov [9] and Balko, Cibulka, Král and Kynčl [1].

2.1. Our bounds. We now give a bound for the oriented Ramsey number of H~ depending on the Ramsey number of H. Our proof will be inspired in the proof of Theorem 2.1 of [3] but, in reality, this idea already appeared in Cochand and Duchet [7] and in Rödl and Winkler [26].

Theorem 2.2. There exists a constant c such that the following holds. Let H~ be an acyclic oriented graph with h vertices and H its underlying undirected graph. There exists orderings <0 and <1 of the vertices of H such that, for H0 = (H,<0) and H1 = (H,<1), we have

c log2(h) R~ (H~ ) 6 R<(H0)+ R<(H1) 6 2R(H) .

4 Proof. Let F~ be the oriented graph formed by two disjoint copies of H~ , in which one has reversed edges. More formally, let F~ be the oriented graph with vertex set

V (F~ ) := V (H~ ) 0, 1 × { } and edge set E(F~ ) := ((u, 0), (v, 0)) , ((v, 1), (u, 1)) : (u, v) E(H~ ) . ∈ Since H~ is acyclic, the orientedn graph F~ is also acyclic. Therefore, there existso an ordering < of the vertices of F~ such that u < v if (u, v) E(F~ ). Let F be the (ordered) underlying undirected ∈ graph of F~ equipped with the ordering <. Let <0 be an ordering of the vertices of H such that, for x,y V (H), we have x< y if and only if (x, 0) < (y, 0). Define < analogously. Let ∈ 0 1 H0 := (H,<0) and H1 := (H,<1). Clearly, we have

R<(F ) 6 R<(H0)+ R<(H1).

Let be an arbitrary ordering of the vertices of KN . We thus consider KN to be an ordered ≺ ord complete graph. By Theorem 2.1, there exists a number N such that K F and N −→ c log2(h) N = R<(F ) 6 R<(H0)+ R<(H1) 6 2R(H) . Now it suffices to prove that K H~ . Let K~ be an arbitrary orientation of K . Color N → N the edges of K in the following way: an edge u, v E(K ) with u v is colored blue if N { } ∈ N ≺ (u, v) E(K~ ) and red otherwise. By the choice of N, there exists an ordered monochromatic ∈ copy of F in K . Let φ : V (F ) V (K ) be the monotone embedding of this copy. If the copy of N → N F in K is blue, then the set of vertices φ((v, 0)) : v V (H~ ) induces a directed copy of H~ in K~ N ∈ with the color blue. Otherwise, if the copyn is red, then the seto of vertices φ((v, 1)) : v V (H~ ) ∈ induces a copy with the color red. In either case we have proved K nH~ , as desired. o N →

Remark 2.3. The proof of Theorem 2.2 shows that the orderings <0 and <1 of V (H~ ) can be taken to be the topological ordering of H~ and the reverse topological ordering of H~ , respectively.

3. A container theorem for digraphs

In preparation for the results of Section 4 and Section 5, we prove a container lemma for digraphs and some supporting lemmas that will be useful in both sections.

3.1. A saturation result for oriented graphs. First we need to prove a saturation result.

Theorem 3.1. Let H~ be an acyclic oriented graph on h vertices, and let R := R~ (H~ ). The following holds for every n > R. For every F E(K ), if there exists an orientation F~ of F ⊆ n such that F~ has at most 2 R −1 n copies of H~ , then h · h

E(K
) F > (1/2R2) n2. | n \ | · R −1 Proof. For convenience, let ε := 2 h . Let F E(Kn) be such that there exists an orien- ~ n ~ ⊆~ tation F of F with at most ε h copies

of H. Let K be an orientation of Kn which agrees with ~ the orientation F of F . Let
V (K~ ) := S : E(K~ [S]) F~ . S ∈ R ⊆ (   )

5 That is, the family is the collection of all R-element subsets S of V (K~ ) such that every edge of S K~ [S] is contained in F~ . By definition of R, every R-element subset of the vertices of K~ contains at least one copy of H~ . This means that, for every S , there exists one copy of H~ in E(K~ [S]). ~ ~ ∈ S n−h Moreover, every copy of H in K is contained in at most R−h R-element subsets. Therefore, double-counting on the pairs (S, H~ ′) where S and H~ ′ is a copy of H~ contained in S yields ∈ S
n−h n h n 1 n 1 n 6 ε − = R−h = . |S| R h h 2 R h 2 R  −   h
    This implies that the set defined as
S V (K~ ) := S R \S   satisfies > (1/2) n . Observe that, by definition of , every set S induces at least one S R S ∈ S edge e E(K~ ) F~ . Moreover, every edge e E(K~ ) F~ is contained in at most n−2 R-element ∈ \
∈ \ R−2 subsets. Now, double-counting on the pairs (S, e) where S and e E(K~ [S]) we get ∈ S ∈
1 n 1 E(K~ ) F~ > S > R > n2. \ n−2 2 n−2 2R2 · R − 2 R−
2

The desired result now follows by observing
E(K ) F
= E(K~ ) F~ .  | n \ | \

3.2. The general container lemma. Let be a l-uniform hypergraph. For a set J V ( ), H ⊆ H we define the degree of J by d(J) := e E( ) : J e . | ∈ H ⊆ | For a vertex v V ( ), we let d(v) := d( v ). For j [l], we also define the maximum j-degree ∈ H { } ∈ of a vertex v V ( ) by ∈ H V ( ) d(j)(v) := max d(J) : v J H . ∈ ∈ j    We denote the average of d(j)(v) for all v V ( ) by ∈ H 1 d := d(j)(v). j v( ) H v∈XV (H) Note that d is the average degree of . Finally, for τ > 0, we define δ as 1 H j dj δj := j−1 d1τ and the co-degree function δ( ,τ) by H l − l −1 − j 1 δ( ,τ) := 2(2) 2 ( 2 )δ . H j Xj=2 We now state a condensed version of the Container Lemma, as expressed in Saxton and Thomason [27]. This version can be found as Theorem 2.1 in [14].

Theorem 3.2 ([27], Corollary 3.6). Let 0 < ε,τ < 1/2. Let = (V, E) be a l-uniform H hypergraph. Suppose that τ satisfies δ( ,τ) 6 ε/(12l!). Then for integers K = 800l(l!)3 and H s = K log(1/ε) the following holds. ⌊ ⌋ 6 For every independent set I V in there exists an s-tuple T = (T ,...,T ) of subsets of ⊆ H 1 s V and a subset C = C(T ) V depending only on S such that ⊆ (a) T I C, i∈[s] i ⊆ ⊆ (b) e(C) 6 ε e( ), and S · H (c) for every i [s] we have T 6 Kτ V . ∈ | i| | | Here we prove a version of the container lemma for H~ -free orientations of graphs. First, we need the following definitions. Definition 3.3. Let H~ be an oriented graph and let n N. Denote by D~ the digraph with ∈ n vertex set [n] and edge set

E(D~ ) := ([n] [n]) (v, v) : v [n] . n × \ { ∈ } We call D~ n the complete digraph. Definition 3.4 ([18], Definition 3.5). Let H~ be an oriented graph with l edges and let n N. ∈ The hypergraph (n, H~ ) = ( , ) is a l-uniform hypergraph with vertex set := E(D~ ) and D V E V n edge set := B V : the edges of B form a digraph isomorphic to H~ . E ∈ l     Definition 3.5. Let H~ be an oriented graph with h vertices. In what follows, we denote by emb := e( (h, H~ )) the number of copies of H~ in D~ . H~ D h 3.3. Checking degree conditions. To apply Theorem 3.2 and prove our container theorems for digraphs, it is first necessary to prove a bound on δ( (n, H~ ),τ) for a suitable value of τ. D This is done by the following lemma.

Lemma 3.6. Let H~ be an oriented graph with h vertices and l > 2 edges. Let also Dτ > 1 and −1/m2(H~ ) write τ := Dτ n . We have

l h−2 −1 δ( (n, H~ ),τ) 6 2(2)h D . D τ Proof. For convenience, set := (n, H~ ). Let J V ( ). Define H D ⊆ H V := a, b [n]. J { }⊆ (a,b[)∈J Note that (V , J) is the subdigraph of D~ induced by the set of edges J. For a set S [n] V such J n ⊆ \ J that S = h V , let emb (J, S) denote the number of copies F~ of H~ such that V (F~ )= V S | | −| J | H~ J ∪ and J E(F~ ). Since emb (J, S) is the same number for any choice of S as above, we write ⊆ H~ embH~ (J) for embH~ (J, S). Note that d(J) is the number of copies of H~ in D~ n which contain the set J. Observe now that n V d(J)= − | J | emb (J). (2) h V H~  − | J | For every j [l], let ∈ f(j) := min v(H′). (3) H~ ′⊆H,e~ (H~ ′)=j It follows from (2) that n V n f(j) d(J)= − | J | emb (J) 6 − emb g(J). h V H~ h f(j) H~  − | J |  −  7 Note now that, for every e V ( ), we have d(1)(e)= d(e)= n−2 emb g( e ). Therefore, the ∈ H h−2 H~ { } average d of all d(1)(e) satisfies d = n−2 emb g( e ), for some fixed e V ( ). It follows 1 1 h−2 H~ { }
∈ H that
n−f(j) n−f(j) f(j)−2 d(J) emb ~ g(J) (h 2)(h 3) . . . (h f(j) + 1) h 6 h−f(j) H 6 h−f(j) 6 n−2 n−2 = − − − . d1 emb
~ g( e )
(n 2)(n 3) . . . (n f(j) + 1) n h−2 H { } h−2 − − −   (j) f(j)−2 2−f(j) Therefore, we have
d (v)/d1 6 h
n . Since f(j) 6 h, this gives us d 1 d(j)(v) 1 j = 6 hf(j)−2n2−f(j) = hf(j)−2n2−f(j) 6 hh−2n2−f(j). d1 v( ) d1 v( ) H v∈XV (H) H v∈XV (H) We furthermore obtain d ~ j 6 h−2 2−f(j) 1−j 6 h−2 2−f(j)+(j−1)/m2(H) 1−j δj = j−1 h n τ h n Dτ . (4) d1τ Observe now that, by definition of m (H~ ), we have m (H~ ) > (j 1)/(f(j) 2). From this we 2 2 − − may derive 2 f(j) + (j 1)/m (H~ ) 6 0. Therefore, we can conclude from (4) that − − 2 h−2 1−j h−2 −1 δj 6 h Dτ 6 h Dτ . (5) Now we can finally bound the co-degree function δ( ,τ) by observing that H l l − − l −1 − j 1 l −1 h−2 −1 − j 1 l h−2 −1 δ( ,τ) = 2(2) 2 ( 2 )δ 6 2(2) h D 2 ( 2 ) 6 2(2)h D . H j τ τ Xj=2 Xj=2 This finishes the proof. 

3.4. A container lemma for graphs with H~ -free orientations. For convenience, given numbers n, s and t, define

(n,s,t) := (T ,...,T ) E(K )s : T 6 t . T  1 s ∈ n i  i∈[s]  [ 

~ We are now able to state and prove our container lemma for grap hs admitting  H-free orientations.

Theorem 3.7. Let H~ be an acyclic oriented graph. There exists a real number α > 0 and positive integers n0, s and c such that the following holds for every n > n0. For every graph G on n vertices such that G H~ , there exists an s-tuple T = (T ,...,T ) E(G)s and a set 6→ 1 s ∈ C = C(T ) E(K ) depending only on T such that ⊆ n (a) i∈[s] Ti E(G) C, ⊆ ⊆2 (b) E(Kn) C > αn , and S| \ | ~ (c) T (n,s,cn2−1/m2(H)). ∈T Proof. We will apply the Container Lemma (Theorem 3.2). Let H~ be an acyclic oriented graph 2 with h vertices and l edges and let R := R~ (H~ ). Let α = 1/(2R ). Let n0 := R and suppose n > n . Set := (n, H~ ) and ε := (2 R emb )−1 and let 0 H D h H~ l h−2
12l!2(2)h D := . τ ε

8 ~ Moreover, let τ := D n−1/m2(H). By Lemma 3.6, this yields δ( ,τ) 6 ε/(12l!). Theorem 3.2 τ H now gives us numbers s and K for , ε and τ. Let G be a graph on n vertices such that G H~ . H 6→ There exists an orientation G~ of G such that G~ contains no copy of H~ . Therefore, the set E(G~ ) is an independent set of . Let T~ = (T~ ,..., T~ ) be an s-tuple of oriented edges and C~ = C~ (T~) H 1 s such as Theorem 3.2 gives for E(G~ ). For i [s], let T be the underlying set of undirected edges ∈ i of T~i. Define C analogously for C~ . By item (a) of Theorem 3.2, we have T E(G) C. i ⊆ ⊆ i[∈[s] We have thus proved item (a). ~ ~ Observe now that embH~ g counts the number of copies of H in any subset of h vertices of Dn, whence it follows that n e( )= emb g. (6) H h H~   ~ R −1 n Therefore, by item (b) of Theorem 3.2 we conclude that C has at most εe( ) = (2 h ) h ~ H · copies of H. By the choice of α, Theorem 3.1 now gives

E(K ) C > αn2. | n \ | We have thus proved item (b). Finally, by letting c := sKDτ , we get by item (c) of Theorem 3.2 that

~ T 6 sKτv( ) 6 cn2−1/m2(H). i H i∈[s] [

We have thus proved item (c). Therefore, there exists an s-tuple T and a set C = C(T ) with the desired requirements. This finishes the proof. 

4. An Oriented Ramsey Theorem for Random Graphs

In this section, we prove the following theorem, applying the results developed in Section 3.

Theorem 4.1. Let H~ be an acyclic oriented graph. There exists a constant C = C(H~ ) such ~ that, if p > Cn−1/m2(H), then lim P G(n,p) H~ = 1. n→∞ → h i Proof. Let α, s, c and n0 be as given by Theorem 3.7 for H~ . Suppose n > n0 and let t := ~ ~ cn2−1/m2(H). Set moreover p := Cn−1/m2(H), for some constant C sufficiently large with respect to c. We will show that P[G(n,p) H~ ]= o(1). 6→ If a graph G on n vertices satisfies G H~ , then by Theorem 3.7 there exists an s-tuple 6→ T = (T ,...,T ) (n,s,t) and a set C(T ) E(K ) such that 1 s ∈T ⊆ n T E(G) C(T ) (7) i ⊆ ⊆ i[∈[s] and E(K ) C(T ) > αn2. | n \ | Let us set for convenience D(T ) := E(K ) C(T ). Since E(G) C(T ), we have n \ ⊆ E(G) D(T )= . (8) ∩ ∅ 9 Let be the family of all graphs G on n vertices such that G H~ . For an s-tuple T = G 6→ (T ,...,T ) (n,s,t), let 1 s ∈T ′ := G = Gn : T E(G) i [s] , GT { i ⊆ ∀ ∈ } and let ′′ := G = Gn : E(G) D(T )= . GT { ∩ ∅} Observations (7) and (8) show that

′ ′′ . G ⊆ GT ∩ GT T ∈T[(n,s,t) As the sets Ti and D(T ) have empty intersection for every i [s], it follows that the events ′ ′′ ∈ [G(n,p) ] and [G(n,p) ] are independent. We conclude ∈ GT ∈ GT ′ ′′ P[G(n,p) ] 6 P G(n,p) P G(n,p) . ∈ G ∈ GT · ∈ GT T ∈TX(n,s,t) h i h i Since D(T ) > αn2 for every T (n,s,t), we have | | ∈T ′′ 2 P G(n,p) 6 (1 p)αn 6 exp( αn2p). ∈ GT − − Moreover, we also have h i

′ P G(n,p) 6 p Si∈[s] Ti . ∈ GT | | T ∈TX(n,s,t) h i T ∈TX(n,s,t) It follows that P[G(n,p) ] 6 exp( αn2p) p Si∈[s] Ti . (9) ∈ G − · | | T ∈TX(n,s,t) We now proceed to bound the sum in (9). For every integer k such that 0 6 k 6 t, define

S(k) := T (n,s,t) : T = k .  ∈T i  i∈[s]  [  n n (2) s k  ( 2)  Observe that S(k) = k (2 ) . Indeed, there are k ways of choosing k edges from E(Kn), s k | | and (2 ) ways of assigning
these edges to the sets of
the s-tuples, which gives the desired equation. Therefore,

t t n t s−1 2 k S T e2 n p p i∈[s] i = S(k) pk 6 2 (2s)kpk 6 1+ . (10) | | | | k k T ∈T (n,s,t) k=0 k=0
k=1 X X X   X   Let f(k) be the function which maps k to (eb/k)k, where b = 2s−1n2p. Observe that f is unimodal and achieves its maximum at k = b. Since 2s−1n2p > cn2p/C = t for C sufficiently

10 large with respect to s and c, we obtain

cn2p/C 2 e2s−1n2p k cn2p Ce2s−1n2p cn p/C 1+ 6 1+ C sufficiently large k C cn2p k=1     X 2 Ce2s−1 cn p/C 6 n2 C sufficiently large c   cn2p = n2 exp (log C +1+(s 1) log 2 log c) C − −   c(log C +1+(s 1) log 2 log c) = n2 exp n2p − − C   α 6 n2 exp n2p C sufficiently large 3 αn 2p  6 exp n sufficiently large. 2   We may now conclude αn2p αn2p P[G(n,p) ] 6 exp( αn2p)exp = exp = o(1), ∈ G − 2 − 2     as desired. 

5. The Isometric Oriented Ramsey Number

5.1. Introduction. Recently, Banakh, Idzik, Pikhurko, Protasov and Pszczoła [3] introduced the concept of isometric oriented Ramsey number, and they proved that the isometric oriented Ramsey number of any acyclic oriented graph is finite. Moreover, they posed the problem of estimating R~ iso(H~ ) for acyclic oriented graphs H~ . In this section, we give an upper bound on R~ iso(H~ ) when H~ is an acyclic orientation of the cycle on k vertices Ck.

Theorem 5.1. There exists a positive constant c1 such that the following holds. Let H~ be an acyclic orientation of Ck and set R := R~ (H~ ). Then

12k3 8k2 R~ iso(H~ ) 6 c1k R . (11)

Remark 5.2. As observed in the Section 1.2, we have R~(H~ ) 6 √2k for every acyclic orientation

H~ of the cycle Ck. In light of Theorem 5.1, one readily sees that there exists a universal constant c1 such that 3 c1k R~ iso(H~ ) 6 k .

The approach employed in this section to prove Theorem 5.1 is very similar to the proof of Theorem 1.1 in Hàn, Retter, Rödl, and Schacht [14]. We will prove a container theorem for graphs with H~ -free orientations, when H~ is an acyclic orientation of a cycle. This will be a more refined version of Theorem 3.7 for this specific case. In particular, we will pay closer attention to the numbers given by the container theorem.

5.2. A container lemma for acyclic orientations of cycles. We begin by observing that, for every orientation H~ of the cycle Ck, we have k 1 m (H~ )= m (C )= − . (12) 2 2 k k 2 − 11 This will justify the choice of constants which we will make in the rest of this section. We now prove the following lemma, which is a slightly improved version of Lemma 3.6 adjusted for orientations of cycles. Our proof makes uses of arguments and results from the proof of Lemma 3.6.

Lemma 5.3. Let H~ be an orientation of the cycle Ck. Let also Dτ > 1 and define τ as 2 −(k−2)/(k−1) (k−1) τ := Dτ n . For every n > Dτ , we have

k k−2 −(k−1) δ( (n, H~ ),τ) 6 2(2)k D . D τ

Proof. Fix j [k]. Let f(j) be as defined in (3). Since H~ is an orientation of the cycle on k ∈ vertices, we have f(j)= j + 1 for every j [k 1] and f(k)= k. Furthermore, by (4) we obtain ∈ − k−2 2−f(j)+(j−1)(k−2)/(k−1) 1−j δj 6 k n Dτ . Therefore, for j [k 1] we have ∈ − k−2 1−j+(j−1)(k−2)/(k−1) 1−j δj 6 k n Dτ k−2 −(j−1)/(k−1) 1−j = k n Dτ k−2 −1/(k−1) −1 6 k k Dτ 6 kk−2n−1/(k−1). (13)

From inequality (5) proved in Lemma 3.6, we obtain that

k−2 −(k−1) δk 6 k Dτ . (14)

(k−1)2 Since, by assumption, we have n > Dτ , inequalities (13) and (14) now give us

max δj = δk. j∈[k] We therefore conclude l l − − l −1 − j 1 l −1 k−2 −(k−1) − j 1 l k−2 −(k−1) δ( ,τ) = 2(2) 2 ( 2 )δ 6 2(2) k D 2 ( 2 ) 6 2(2)k D , H j τ τ Xj=2 Xj=2 as desired. 

We are now able to state and prove our container lemma for graphs admitting H~ -free orien- tations in the specific case when H~ is an acyclic orientation of Ck.

Theorem 5.4. Let H~ be an acyclic orientation of Ck. There exists positive integers n0, s and c such that the following holds for every n > n0. For every graph G on n vertices such that G H~ , there exists an s-tuple T = (T ,...,T ) E(G)s and a set C = C(T ) E(K ) 6→ 1 s ∈ ⊆ n depending only on T such that

(a) T E(G) C, i∈[s] i ⊆ ⊆ (b) E(K ) C > n2/(2R2), and S| n \ | (c) T (n,s,cn2−(k−2)/(k−1)). ∈T 12 In particular, the constants can be chosen to be the following:

(k−1)2 n0 = D , (15) s = K log(2Rk) 6 1600k3k+2R, (16) ⌊ ⌋ c = sKD, (17) where R = R~ (H~ ) and

D = 4 2k/2 k2 (2Rk)1/(k−1) 6 8R2kk+2, (18) · · · K = 800k(k!)3 6 800k3k+1. (19)

k Proof. The proof is very similar to Theorem 3.7. Let ε := 1/(2R ) and suppose n > n0. Observe that k−1 k k 4 2k/2 k2 4k−1 2(2) k2(k−1) 12 2(2) kk−2 k! Dk−1 = · · = · · > · · · . ε1/(k−1) ! ε ε Moreover, let τ = Dn−(k−2)/(k−1). Lemma 5.3 now yields δ( (n, H~ ),τ) 6 ε/(12k!). Theorem 3.2 D gives us numbers s and K for , ε and τ, where H K = 800k(k!)3, (20) s = K log(1/ε) = K log(2Rk) . (21) ⌊ ⌋ ⌊ ⌋ Now item (a) can be proved in exactly the same way as in the proof of Theorem 3.7. Item (b) can proved by observing that 1 1 1 ε = 6 6 . 2Rk R R 2k! k 2 embH~ g k n 6 Therefore, every container C will admit an orientation
which
has at most ε embH~ g k R −1 n ~ (2 k ) k copies of H. Item (b) then follows from Theorem 3.1. Finally, item (c) can
be  shown
just
as in the proof of Theorem 3.7, by letting c := sKD.

5.3. Proof of Theorem 5.1. We may now proceed to the proof of Theorem 5.1. The proof will be as follows. We will consider the random graph G(n,p) and, imitating the proof of iso Theorem 4.1, we will prove that, with positive probability, we have G(n,p) H~ , for a number −→ n that satisfies (11) and a suitable choice of p. Our strategy will be to prove that the graph

G(n,p) has girth at least k and satisfies G(n,p) H~ for an acyclic orientation H~ of Ck, which iso → implies G(n,p) H~ . −→

Proof of Theorem 5.1. Let n0, s ,c, D and K be as given by Theorem 5.4 for H~ . We begin by setting the following numbers we are going to use in the proof:

2 2 2 Dp = KDs 10R log(5R ), (22) k2 n = Dp , (23) k−2 − − p = Dpn k 1 . (24)

Observe that, for some positive constant c1 > 0, we have D 6 c k10k+7R8 6 k12kR8, p 1 · 13 which implies 12k3 8k2 n 6 c2k R .

Moreover, we also have n > n0. Therefore, we can apply Theorem 5.4 to graphs with n vertices. Let us first prove the following claim. The proof goes just as in the proof of Claim 3.1 of [14].

Claim 5.5. We have P[girth(G(n,p)) > k] > exp( kDk−1n). − p

Proof. Let (n, k) be the set of all cycles C E(K ) of length at most k 1. Let C ⊆ n − X := C (n, k) : C E(G(n,p)) |{ ∈C ⊆ }| be the random variable counting the number of cycles of length at most k 1 in G(n,p). For − each cycle C E(K ) of length at most k 1, let X be the indicator function of the event ⊆ n − C E := C E(G(n,p)) . Clearly, X is the sum of all such C. Therefore, C { ⊆ } k−1 k−1 (j 1)! n (pn)j k k E[X]= p|C| = − pj 6 6 (pn)k−1 = Dk−1n. 2 j 2j 6 6 p C∈CX(n,k) Xj=3   Xj=3 Moreover, the set of all graphs G on n vertices such that C E(G) is a monotone decreasing 6⊆ property. Therefore, using the FKG inequality and applying the inequality 1 x > exp( x/(1 − − − x)) for x [0, 1), we get ∈ p|C| E[X] P > |C| > > [girth(G(n,p)) k]= (1 p ) exp exp 3 . − −1 p|C| ! −1 p C∈CY(n,k) C∈CY(n,k) −  −  One may now easily check that

2 1 p3 = 1 n−(k−2)/(k−1)+1/k > 1/6, − − since n> 11, and the claim follows. 

We now prove the following claim. Our proof will be similar to that of Theorem 4.1, with the difference that the calculations will be more involved.

Claim 5.6. We have n2p P[G(n,p) H~ ] > 1 exp . → − −4R2   Proof. Applying Theorem 5.4 instead of Theorem 3.7, we can follow the proof of Theorem 4.1 up until inequalities (9) and (10). We then get

t n2p e2s−1n2p j P[G(n,p) H~ ] 6 exp 1+ , (25) 6→ −2R2  j    Xj=1     where t = sKDn2−(k−1)/(k−2). We now proceed to bound the sum in (25). Let f(k) be the function which maps j to (eb/j)j , where b = 2s−1n2p. Observe that f is unimodal and achieves its maximum at k = b. Observe moreover that

k−2 k−2 s−1 2 s−1 − − 2 − − 2 2 n p = 2 Dpn k 1 n > sKDn k 1 n .

14 whence it follows that t e2s−1n2p k e2s−1n2p t e2s−1D t 1+ 6 1+ t =1+ t p . k t sKD Xk=1       Moreover, since

−(k−2)/(k−1) −k2(k−2)/(k−1) −1 sKDn = sKDDp 6 sKDDp < 1, we obtain e2s−1D t e2s−1D t 1+ t p 6 n2 p sKD sKD     e2s−1D = n2 exp t log p · sKD   sKD e2s−1D = n2 exp n2p log p · D · sKD  p  sKD D = n2 exp n2p log(e2s−1) + log p . · D sKD p   ! Observe now that sKD log(e2s−1) 1 1 s−1 6 6 log(e2 )= 2 2 2 2 2 . (26) Dp s10R log(5R ) 10R log(5R ) 10R

Let now x := Dp/(sKD) and set y := x/s. Since the function log(x)/x is decreasing for x > e, we have log(x)/x 6 log(y)/y. Note also that y = 10R2 log(5R2) 6 (5R2)2, since log x < x/2 for x> 0. applying this inequality once again, we obtain log y 6 log(5R2). These observations allow us to conclude that log(D /(sKD)) log x log y log(5R2) 1 p 6 6 = 2 2 = 2 . (27) Dp/(sKD) x y 10R log(5R ) 10R Hence, by inequalities (26) and (27) we obtain

sKD D n2p n2 exp n2p log(e2s−1) + log p 6 n2 exp . · D sKD 5R2 p   !   Observe now that 2 2 2k2 −k (k−2)/(k−1) 2k2 −k2 k2 2 n p D DpDp D DpD D D 10R D p > p p p > p > > 2 = 2 2 = 2 2 2 40R . log n k log(Dp) k Dp k k k From this we obtain n2p n2p n2p 2 log n 6 = , 20R2 4R2 − 5R2 which implies n2p n2p n2 exp 6 exp . 5R2 4R2     All our work so far therefore implies t e2s−1n2p j n2p 1+ 6 exp , j 4R2 Xj=1    

15 which, in view of (25), yields n2p P[G(n,p) H~ ] 6 exp . 6→ −4R2   This finishes the proof of the claim. 

Now, in view of Claim 5.5 and Claim 5.6, we can deduce P[girth G(n,p) > k and G(n,p) H~ ] > P[girth(G(n,p)) > k]+ P[G(n,p) H~ ] 1 → → − n2p (28) > exp( kDk−1n) exp . − p − −4R2   Since we also have 2 2 n2p Dk +1−k (1−1/(k−1)) Dk+1 = n p >n p > kDk−1n, 4R2 · 4R2 · 4R2 p we may now conclude from (28) that

P[girth G(n,p) > k and G(n,p) H~ ] > 0, → which finishes the proof. 

6. Concluding remarks

The landscape of Ramsey theory for oriented graphs is much less explored than for undirected graphs. It would be interesting to obtain better bounds for the oriented Ramsey number of specific families of graphs. In particular, we leave unaddressed the question of whether the upper bound of Theorem 2.2 is tight for some oriented graph. Moreover, one could also consider not only orientations of graphs, but also orientations and colorings of edges, and require the oriented copy to be monochromatic. The resulting Ramsey number was studied, for instance, in a work due to Bucić, Letzter and Sudakov [6]. We believe that an upper bound to the threshold probability of this property, like that of Theorem 4.1, can also be proved with the techniques of this work. Our work also leaves open the question of proving a lower bound for the threshold probability of the property G(n,p) H~ , matching the upper bound of Theorem 4.1. One can easily check { → } that this upper bound is not tight for the transitive tournament on 3 vertices (if G contains a K4, then every orientation of G contains the tournament on 3 vertices). It would be very interesting to understand for which oriented graphs this upper bound is tight, and to prove tighter bounds when this is not the case. Finally, one could also try to apply the techniques of Section 5 to derive bounds for the isometric Ramsey number of other graphs, like directed paths.

7. Acknowledgements

The author thanks Marcelo Tadeu de Sá Oliveira Sales for helpful comments in an early stage of this work, and Yoshiharu Kohayakawa and Tássio Naia for proofreading an earlier version of this work and for providing numerous suggestions, helpful comments and encouragements.

16 References

[1] Martin Balko, Josef Cibulka, Karel Král, and Jan Kynčl. Ramsey numbers of ordered graphs. Electronic Notes in Discrete Mathematics, 49:419 – 424, 2015. The Eight European Conference on Combinatorics, Graph Theory and Applications, EuroComb 2015. 1.2, 2, 2 [2] József Balogh, Robert Morris, and Wojciech Samotij. Independent sets in hypergraphs. J. Amer. Math. Soc., 28(3):669–709, 2015. 1.3 [3] Taras Banakh, Adam Idzik, Oleg Pikhurko, Igor Protasov, and Krzysztof Pszczoła. Isometric copies of directed trees in orientations of graphs. Journal of Graph Theory, pages 1–17, 2019. 1.4, 2.1, 5.1 [4] Gary S. Bloom and Stefan A. Burr. On unavoidable digraphs in orientations of graphs. J. Graph Theory, 11(4):453–462, 1987. 1.2 [5] J. A. Bondy and U. S. R. Murty. Graph theory, volume 244 of Graduate Texts in Mathematics. Springer, New York, 2008. 1.2 [6] Matija Bucić, Shoham Letzter, and Benny Sudakov. Directed Ramsey number for trees. J. Combin. Theory Ser. B, 137:145–177, 2019. 6 [7] M. Cochand and P. Duchet. A few remarks on orientation of graphs and Ramsey theory. In Irregularities of partitions (Fertőd, 1986), volume 8 of Algorithms Combin. Study Res. Texts, pages 39–46. Springer, Berlin, 1989. 2.1 [8] D. Conlon, D. Dellamonica, Jr., S. La Fleur, V. Rödl, and M. Schacht. A note on induced Ramsey numbers. arXiv e-prints, June 2016. 1.3 [9] David Conlon, Jacob Fox, Choongbum Lee, and Benny Sudakov. Ordered Ramsey numbers. J. Combin. Theory Ser. B, 122:353–383, 2017. 1.2, 2, 2.1, 2 [10] A. El Sahili. Trees in tournaments. J. Combin. Theory Ser. B, 92(1):183–187, 2004. 1.2 [11] P. Erdős and L. Moser. On the representation of directed graphs as unions of orderings. Magyar Tud. Akad. Mat. Kutató Int. Közl., 9:125–132, 1964. 1.1 [12] P. Erdös and G. Szekeres. A combinatorial problem in geometry. Compositio Math., 2:463–470, 1935. 1 [13] Roland Häggkvist and Andrew Thomason. Trees in tournaments. Combinatorica, 11(2):123–130, 1991. 1.2 [14] H. Hàn, T. Retter, V. Rödl, and M. Schacht. Ramsey-type numbers involving graphs and hypergraphs with large girth. Combin. Probab. Comput. To appear. 1.3, 1.4, 3.2, 5.1, 5.3 [15] Frédéric Havet and Stéphan Thomassé. Oriented Hamiltonian paths in tournaments: a proof of Rosenfeld’s conjecture. J. Combin. Theory Ser. B, 78(2):243–273, 2000. 1.2 [16] M.-C. Heydemann, D. Sotteau, and C. Thomassen. Orientations of Hamiltonian cycles in digraphs. Ars Combin., 14:3–8, 1982. 1.2 [17] Daniela Kühn, Richard Mycroft, and Deryk Osthus. A proof of Sumner’s universal tournament conjecture for large tournaments. Proc. Lond. Math. Soc. (3), 102(4):731–766, 2011. 1.2 [18] Daniela Kühn, Deryk Osthus, Timothy Townsend, and Yi Zhao. On the structure of oriented graphs and digraphs with forbidden tournaments or cycles. J. Combin. Theory Ser. B, 124:88–127, 2017. 3.4 [19] Richard Mycroft and Tássio Naia. Unavoidable trees in tournaments. Random Structures Algorithms, 53(2):352–385, 2018. 1.2 [20] Tássio Naia. Large Structures in Dense Directed Graphs. PhD thesis, University of Birmingham, 7 2018. 1.2 [21] Rajko Nenadov and Angelika Steger. A short proof of the random Ramsey theorem. Combin. Probab. Com- put., 25(1):130–144, 2016. 1.3 [22] Stanisław P. Radziszowski. Small Ramsey numbers. Electron. J. Combin., 1:Dynamic Survey 1, 30, 1994. 1 [23] F. P. Ramsey. On a problem of formal logic. Proc. London Math. Soc. (2), 30(4):264–286, 1929. 1 [24] Vojtěch Rödl and Andrzej Ruciński. Threshold functions for Ramsey properties. J. Amer. Math. Soc., 8(4):917–942, 1995. 1, 1.3, 1.2 [25] Vojtěch Rödl, Andrzej Ruciński, and Mathias Schacht. An exponential-type upper bound for folkman num- bers. Combinatorica, May 2016. 1.3 [26] Vojtěch Rödl and Peter Winkler. A Ramsey-type theorem for orderings of a graph. SIAM J. Discrete Math., 2(3):402–406, 1989. 2.1 [27] David Saxton and Andrew Thomason. Hypergraph containers. Invent. Math., 201(3):925–992, 2015. 1.3, 3.2, 3.2

17 [28] Andrew Thomason. Paths and cycles in tournaments. Trans. Amer. Math. Soc., 296(1):167–180, 1986. 1.2

Instituto de Matemática e Estatística, Universidade de São Paulo, Rua do Matão 1010, 05508– 090 São Paulo, SP Email addresses: [email protected]

18