NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES

Transcription

1 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES XING SHI CAI AND SVANTE JANSON Abstract. We study ST n, the number of subtrees in a conditioned Galton Watson tree of size n. With two very different methods, we show that logst n has a Central Limit Law and that the moments of ST n are of exponential scale. We define the model which we study in Section 1. Our main results are given in Section 2; the proofs can be found in Sections 3, 4 and 5 respectively. An extension is given in Section Definitions 1.1. Subtrees. We consider only rooted trees. We denote the node set of a rooted tree T by V T, and the number of nodes by T = V T. We denote the root of T by o = ot. We regard the edges of a rooted tree as directed away from the root. A general subtree of a rooted tree T is a subgraph T that is a tree. T is necessarily an induced subgraph, so we may identify it with its node set V = V T ; hence we can also define a subtree as any set of nodes that forms a tree; in other words, any non-empty connected subset V of the node set V T. Note that a subtree T of T has a unique node o of smallest depth in T, and that all edges in T are directed away from o. We define o to be the root of T. Thus every subtree T is itself a rooted tree, with the direction of any edge agreeing with the direction in T. A fringe subtree is a subtree T that contains all children of any node in it, i.e., if v V = V T then w V for every child w of v. Equivalently, a fringe subtree is the tree T v consisting of all descendants in T of some node v V T which becomes the root of T v. Hence the number of fringe subtrees of T equals the number of nodes of T. Fringe subtrees are studied in many papers; often they are simply called subtrees. To avoid confusion, we call the general subtrees studied in the present paper non-fringe subtrees. This is a minor abuse of notation, since fringe subtrees are examples of non-fringe subtrees; the name should be interpreted as not necessarily fringe. A root subtree of a rooted tree T is non-fringe subtree T that contains the root ot which then becomes the root of T too. Equivalently, a root subtree is a non-empty set V V T such that if v V, then the parent of v also belongs to V. Date: 6 March Partly supported by the Knut and Alice Wallenberg Foundation. 1

2 2 XING SHI CAI AND SVANTE JANSON Let ST be the set of non-fringe subtrees of T, and RT the subset of root subtrees. Let ST := ST be the number of non-fringe subtrees of T, and RT := RT the number of root subtrees. Note that a non-fringe subtree of T is a root subtree of a unique fringe subtree T v. Hence, ST = v T RT v. 1.1 Furthermore, for any v T, RT v RT, since we obtain an injective map RT v RT by adding to each tree T RT v the unique path from o to v. Consequently, using 1.1, RT ST T RT, Conditioned Galton Watson trees. A Galton Watson tree T is a tree in which each node is given a random number of child nodes, where the numbers of child nodes are drawn independently from the same distribution ξ which is often called the offspring distribution. We use ξ to denote both the offspring distribution and a random variable with this distribution. Galton Watson trees were implicitly introduced by Bienaymé [1] and Watson and Galton [10] for modeling the evolution of populations. A conditioned Galton Watson tree T n is a Galton Watson tree conditioned on having size n. It is well-known that T n encompasses many random tree models. For example, if P ξ = i = 2 i 1, i.e., ξ has geometric 1/2 distribution, then T n is a uniform random tree of size n. Similarly, if P ξ = 0 = P ξ = 2 = 1/2, then T n is a uniform random full binary tree of size n. As a result, the properties of T n has been well-studied. See, e.g., [7] and the references there. For fringe and non-fringe subtrees of T n, see [8; 4; 2; 3] Simply generated trees. Let w i i 0 be a given sequence of nonnegative numbers, with w 0 > 0. For a tree T, let D + v be the out-degree number of children of a node v T, and define the weight of T by wt = v T w D+ v. 1.3 Let T n [s] be a tree chosen at random from all ordered trees of size n with probability proportional to their weights. In other words, P T n [s] wt = T = 1.4 wt. T : T =n We call T n [s] a simply generated tree with weight sequences w i i 0, and the generating function Φz := w i z i. 1.5 i 0 its generator. Note that the conditioned Galton Watson tree T n with the offspring distribution ξ is the same as the T n [s] with the weight sequence P ξ = i i 0. In this case, the generator Φz is just the probability generating function of ξ. Hence, simply generated trees generalize conditioned Galton Watson

3 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 3 trees. On the other hand, given a sequence w i with generator Φz, any sequence with a generator aφbz with a, b > 0 yields the same T n [s], and in many cases a and b can be chosen such that the new generator is a probability generating function, and then T n [s] is a conditioned Galton Watson tree. Consequently, simply generated trees and conditioned Galton Watson trees are essentially the same, and we use in the sequel the notation T n for both. In particular, see, e.g., [7, Section 4], a simply generated tree with generator Φz is equivalent to a conditioned Galton Watson tree with offspring distribution ξ satisfying E ξ = 1 and E e tξ < for some t > 0, if and only if Φz has a positive radius of convergence R 0, ] and zφ z lim > z R Φz Although the two formulations are equivalent under our conditions, the formulation with simply generated trees is sometimes more convenient, since it gives more flexibility in choosing a convenient Φ; see for example Section 4.1. For more on the connection between the two models, see [6, pp ] and [7, Sections 2 and 4] Some further notation. If v and w are nodes in a tree T, then v w denotes that v is ancestor of w. We denote T being a non-fringe general subtree of T by T T and T being a root subtree of T by T r T. For a formal power series fz := n f nz n, we let [z n ]fz := f n. 2. Main results We give two types of results in this paper, proved by two different methods. First, both RT n and ST n have an asymptotic log-normal distribution, as conjectured by Luc Devroye personal communication. Theorem 2.1. Let T n be a random conditioned Galton Watson tree of order n, defined by some offspring distribution ξ with E ξ = 1 and 0 < Var ξ <. Then there exist constants µ, σ 2 > 0 such that, as n, log RT n µn n log ST n µn n d N0, σ 2, 2.1 d N0, σ 2, 2.2 where N0, σ 2 denotes the normal distribution with mean 0 and variance σ 2. Furthermore, E[log RT n ] = E[log ST n ] + Olog n = nµ + o n, 2.3 Var[log RT n ] = Var[log ST n ] + on = nσ 2 + on. 2.4 The proof is given in Section 3, and is based on a general theorem in [8]. It is in principle possible to calculate µ and σ 2 in Theorem 2.1, at least numerically, see Remark 3.5. Secondly, if we also assume that ξ has a finite exponential moment a mild assumption satisfied by all standard examples, then we can use generating

4 4 XING SHI CAI AND SVANTE JANSON functions and singularity analysis to obtain asymptotics for the mean and higher moments of RT n. Theorem 2.2. Let T n be as in Theorem 2.1, and assume further that E e tξ < for some t > 0. Assume further that if R is the radius of convergence of the probability generating function Φz := E z ξ, then Φ R := lim z R Φ z =. Then there exist sequences of numbers γ m > 0 and 1 < τ 1 < τ 2 <... such that for any fixed m 1, E RT n m = 1 + On 1 γ m τ n m. 2.5 We will later use the formulation of simply generated trees. In this language, Theorem 2.2 has the following, equivalent, formulation. Theorem 2.3. Let T n be a simply generated tree with generator Φz. Let R be the radius of convergence of Φz. Assume that R > 0 and that Then 2.5 holds. zφ z lim > 1, 2.6 z R Φz Φ R := lim z R Φ z =. 2.7 The proof of Theorems is given in Section 4. We first Sections illustrate the argument by studying the simple case of full binary trees, where we do explicit calculations. Similar explicit calculations could presumably be performed, e.g., for full d-ary trees, or for ordered trees. Then we give the proof for the general case in Section 4.3. Note that the condition 2.6 is the same as 1.6; however, we need also the extra condition 2.7. The latter condition is a weak assumption that is satisfied in most applications, and in particular if R =, or if ΦR =. Nevertheless, this extra condition or some other is necessary; we give in Section 4.4 an example showing that Theorems are not valid without 2.7. For moments of ST n, we have by 1.2 the same exponential growth τm, n but possibly also a polynomial factor. In fact, there is no such polynomial factor, and E ST n m and E RT n m differ asyptotically only by a constant factor, as shown by the following theorem, proved in Section 5. Theorem 2.4. Let T n be as in Theorem 2.2 or 2.3. Then, for any m 1, where τ m is as in 2.5 and γ m > 0. More generally, for m, l 0, for some γ m,l > 0. E ST n m = 1 + On 1 γ mτ n m, 2.8 E[RT n l ST n m ] = 1 + On 1 γ m,l τ n l+m, 2.9 The constants γ m,l can be calculated explicitly, see Remark 2.5. We can express 2.1 and 2.2 by saying that RT n and ST n have the asymptotic distribution LNnµ, nσ 2. Note that if W LNnµ, nσ 2 exactly, so W = e Z with Z Nnµ, nσ 2, then the moments of W are given by E W m = E e mz = e mnµ+m2 nσ 2 /2 = e mµ+m2 σ 2 /2n. 2.10

5 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 5 We may compare this to Theorem 2.2 and ask whether τ m? = e mµ+m 2 σ 2 / It seems natural to guess that equality holds in 2.11; however, we show in Remark 4.3 that it does not hold, at least not for all m, in the case of full binary trees. We therefore conjecture that, in fact, equality never holds in This may seem surprising; however, note that the same happens in the simpler case Y = e X with X Bin, p, with p fixed. Then Y is asymptotically LNnp, np1 p in the sense above, but E Y m = E e mx = 1 + pe m 1 n while if W LNnp, np1 p, then E W m = e mp+m2p1 p/2n, with a different basis for the n:th power. 3. Proof of Theorem 2.1 Proof of Theorem 2.1. First, by 1.2, log ST n log RT n log n, and thus 2.1 and 2.2 are equivalent. Similarly, the first inequalities in 2.3 and 2.4 hold, using also Minkowski s inequality for the latter. We consider in the rest of the proof only RT n. Suppose that the root o of T has D children v 1,..., v D, and write T i := T vi. Then, a root subtree of T consists of the root o and, for each child v i, either the empty set or a root subtree of T i. Consequently, D RT = RTi i=1 Define F T := log RT + 1 = log RT + O Then 3.1 implies F T = log RT + log 1 + RT 1 D = F T i + log 1 + RT In other words, F T is an additive functional with toll function ft := log 1 + RT 1, see e.g. [8, 1]. For any tree T, and any node v T, the path from the root o to v is a root subtree. Hence, RT T, 3.4 and as a consequence, i=1 0 ft := log 1 + RT 1 RT 1 T In particular, we have the deterministic bound ft n 1/n. This bound implies that the conditions of [8, Theorem 1.5] are satisfied, and that theorem, together with the estimate in 3.2, yields 2.1, 2.3 and 2.4, for some µ, σ 2 0. Furthermore, if T is the unconditioned Galton Watson tree with offspring distribution ξ, then µ = E ft > It remains only to verify that σ 2 > 0. This is expected in all applications of [8, Theorem 1.5], except trivial ones where F T n is deterministic for all large n, but we do not know any general result; cf. [8, Remark 1.7]. In the present case, it can be verified as follows.

6 6 XING SHI CAI AND SVANTE JANSON Consider a tree T. Denote the depth and out-degree number of children of a node v T by dv and D + v. Fix a node v T, write d = dv, and let the path from o to v be o = v 0, v 1,..., v d = v. By 3.1, we have for j = 0,..., d 1, RT vj = α j RTvj+1 + 1, 3.7 where α j is the product of RT w + 1 over all children w v j+1 of v j. Note that each RT w 1, and thus Define and γv := α j 2 D +v j 1 D + v j. 3.8 d j=1 d 1 βv := α j, 3.9 βv j βv = j=0 d d 1 αv k j=1 k=j Then repeated applications of 3.7 i.e., induction on d yield the expansion RT = RT v0 = Hence, with we have Define also d βv j + βvrt v = βv RT v + γv j=1 γ v := γv + βv 1 = d j=0 βv j βv, 3.12 F T = log RT + 1 = log βv + log RT v + γ v γ v := d d 1 D + v k 1, 3.14 j=0 k=j and note that γ v γ v by Now, let T be a modification of T, where the subtree T v is replaced by some tree T v, but all other parts of T are left intact. Then all α j, βv j, γv, γ v and γ v are the same for T as for T. Hence, if we further assume that RT v < RT v, then 3.13 yields F T F T = log RT v + γ v log RT v + γ v log RT v + γ v log RT v 1 + γ v RT v + γ v 1 RTv + γ v Next, fix an l 2 be such that Pξ = l > 0. Let T a be a tree where the root o and two of its children have out-degrees l, and all other nodes have out-degree 0 i.e., they are leaves. Similarly, let T b be a tree where o, one of its children, and one of its grandchildren have out-degree l, and all other nodes have out-degree 0. Then both T a and T b are trees of order 3l + 1, and

7 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 7 both are attained with positive probability by T 3l+1. Furthermore, a simple calculation using 3.1 shows that RT a = 2 l 2 2 l = 2 3l l l 2, 3.16 RT b = 2 l 1 2 l 1 2 l = 2 3l l l 1, 3.17 and thus RT a > RT b. Consequently, the random variable RT 3l+1 is not a.e. equal to a constant. Fix also a large constant A, to be chosen later, and say that a node v T is good if T v = 3l+1 and γ v A. Define the core T of T as the subtree obtained by marking all good nodes in T and then deleting all descendants of them. Note that adding back arbitrary trees of order 3l + 1 at each marked node of T yield a tree T of the same order as T, and with the same good nodes, because T v and γ v are unchanged for every v T. It follows that the random tree T n, conditioned on its core Tn = T, consists of T with an added tree T v at each good i.e., marked node of T, and that these added trees T v all have order 3l + 1 and are independent copies of T 3l+1. Now suppose in order to obtain a contradiction that σ 2 = 0; then 2.1 and 3.2 show that F T n µn/ n p 0. In particular, P F T n nµ > n We show in Lemma 3.1 below that there exists a constant c > 0 such that, for large n, T n has with probability 1/2 at least cn good nodes. Hence, 3.18 holds also if we condition on the existence of at least cn good nodes. Condition further on the core Tn, and among the possible cores T of T n with at least cn good nodes, choose one that minimizes P F T n nµ > n T n = T. For each n, fix this choice T = T n, and note that P F T n nµ > n T n = T P F T n nµ > n at least cn good nodes Let m be the number of good i.e., marked nodes in T = T n and label these v 1,..., v m. Condition on Tn = T. Then, as noted above, T n consists of T with a tree T i added at v i, for each i, and these trees T 1,..., T m are m independent copies of T 3l+1. Let X i := RT i ; thus X 1,..., X m are i.i.d. random variables with some fixed distribution. Furthermore, repeated applications of 3.1 show that RT n is a function depending on T n of X 1,..., X m. Hence, by 3.2, we have, still conditioning on Tn = T, F T n = g n X 1,..., X m, 3.20 for some function g n. Consequently, writing Y m := g n X 1,..., X m, we have by 3.19 P Y m nµ > n = P F T n nµ > n T n = T 0, 3.21 as n. Recalling that m cn, this implies P Y m nµ > c 1/2 m We now obtain the sought contradiction from 3.40 in Lemma 3.4 below. To be precise, we use a relabelling. We have m = mn as n ; we may select a subsequence with increasing m and consider this sequence only, relabelling g n as g m. Note that in this application of Lemma 3.4, S is

8 8 XING SHI CAI AND SVANTE JANSON a finite set of integers the range of RT 3l+1. The conditions of Lemma 3.4 are satisfied: by , we can find s such that 0 < PX 1 s < 1; furthermore, 3.39 holds under the stated condition with δ := 2 3l+1 + A 1 by 3.15, since γ v i A by the definition of good vertices and RT v 2 Tv = 2 3l 1. This completes the proof that σ 2 > 0, given the lemmas below. Lemma 3.1. With notations as above, there exists A < and c > 0 such that, for large n, P T n has at least cn good nodes 1/2. Proof. Note first that if Pξ = 1 = 0, and thus T n has no nodes of out-degree 1, then this is easy. In this case, 3.14 yields γ v d j=0 2j d < 2 for every v, since D + v k 2 for each v k. Taking A = 2, every node v with T v = 3l + 1 is thus good. If n 3l+1 T n denotes the number of these nodes in T n, then and thus n 3l+1 T n /n p P T = 3l + 1 > 0, as n, 3.23 P n 3l+1 T n > cn for any c < P T = 3l + 1. In general, 3.24 still holds, but there is no uniform bound on γ v, as is shown by the case of a long path, and it remains to show that γ v is bounded for sufficiently many nodes. We define, for a given tree T and any pair of nodes v, w with v w, πu, v := D + w We then can rewrite 3.14 as u w v γ v = u v πu, v, 3.26 and we define also the dual sum ζv = w v πv, w Note that ζv is a functional of the fringe subtree T v. We write ζt := ζo, where o is the root of T ; then for an arbitrary node v T, ζv = ζt v. We may also note, although we do not use this explicitly, that ζv has a natural interpretation: πv, w is the probability that a random walk, started at v and at each step choosing a child uniformly at random, will pass through w. Hence, ζv is the expected length of this random walk. If the root o of T has D children v 1,..., v D, and the corresponding fringe trees are denoted T 1,..., T D, then ζt = w T πo, w = 1+ D i=1 w T i D 1 πv D, w = 1+D 1 D ζt i We apply this with T = T n, the conditioned Galton Watson tree. Note that conditioned on the root degree D, and the sizes n i := T i of the subtrees, i=1

9 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 9 each T i is a conditioned Galton Watson tree T ni. Consequently, 3.28 yields We claim that E ζt n D, n 1,..., n D = 1 + D 1 D E ζt ni i=1 E ζt n C for some constant C 1 and all n. We prove this by induction, assuming that 3.30 holds for all smaller n. Note also that if T = 1, then ζt = 1. Hence, by 3.29 and the induction hypothesis, if D 1 := {i : n i = 1}, the number of children of o that are leaves, then E ζt n D, n 1,..., n D 1 + D 1 D D 1 C 1 + D 1 and hence = C D 1 C 1 1/D E ζt n C C 1 1 ED 1 /D, 3.32 where D 1 and D are calculated for the random tree T n. As n, the distribution of the pair D, D 1 converges to the ˆD, ˆD 1, the same quantities for the random limiting infinite tree ˆT, see for example [7, Section 5 and Theorem 7.1]. Hence, using bounded convergence, ED 1 /D E ˆD 1 / ˆD > 0 as n. Since PD 1 > 0 > 0 for every n, and thus ED 1 /D > 0 for every n, it follows that there exists a constant c 1 > 0 such that for every n, ED 1 /D c 1. If we choose C 1 = 1 + 1/c 1, then 3.32 yields E ζt n C 1, which verifies the induction step. Hence, 3.30 holds for all n. Next, let, for any tree T, ZT := v T ζt v, 3.33 the additive functional with toll function ζt. It follows from 3.30 that E ζt = n 1 P T = n E [ζt n ] C 1, 3.34 where T denotes an unconditioned Galton Watson tree. By [8, Remark 5.3], it follows from 3.34 and 3.30 that E ZT n n E ζt = On Thus there exists a constant C 2 such that for all n 1, E ZT n C 2 n Consequently, by Markov s inequality, with probability 2 3, ZT n 3C 2 n For any tree T, 3.33 and yield ZT = ζv = πv, w = γ w v T v,w:v w w T Hence, if we choose A := 6C 2 /c, then 3.37 implies that at most 3C 2 n/a = cn/2 nodes w in T n satisfy γ w > A, and hence at least n 3l+1 T n cn/2

10 10 XING SHI CAI AND SVANTE JANSON nodes are good. This and 3.24 show that, with probability o1, T n has at least cn/2 good nodes. Remark 3.2. As the proof shows, the probability 1/2 in Lemma 3.1 can be replaced by any number < 1. We conjecture that in fact, for suitable A and c, the probability tends to 1. Remark 3.3. If we assume that the offspring distribution ξ has an exponential moment, so that its probability generating function has radius of convergence > 1, then one can alternatively derive 3.30 and 3.36, and precise asymptotics, using generating functions. We leave this to the reader. Lemma 3.4. Let X 1, X 2,... be i.i.d. random variables, with values in some set S R. Let Y m = g m X 1,..., X m, for some functions g m : S m R, m 1, and assume that there is a number s and a δ > 0 such that 0 < PX 1 s < 1 and that g m y1,..., y j 1, y j, y j+1,..., y m gm y1,..., y j 1, y j, y j+1,..., y m + δ, 3.39 for any m, j m, y 1,..., y m S and y j S, such that y j s and y j > s. Then, for any constant B and any sequence µ m, lim sup P Y m µ m B m < m Proof. First, by replacing g m by g m µ m, we may assume that µ m = 0. If 3.40 does not hold, then, by restricting attention to a subsequence, we may assume P Y m B m 1, as m. Let N m := {i : X i > s}. Thus N m has a binomial distribution Bim, p, where p := PX 1 > s 0, 1. Fix a large number K > 0, and define the events E m + := {N m > mp + K m} and Em := {N m < mp K m}. By the central limit theorem for the binomial distribution, PE m + q and PEm q for some q > 0, and thus our assumption implies that P Y m B m E + m 1, P Ym B m E m Hence we can find integers n + m and n m with 0 n m < mp K m < mp + K m < n + m n such that P Y m B m N m = n + m 1, P Ym B m N m = n m Choose e.g. n ± m as the integers in the allowed ranges that maximize these probabilities. Let Xm = X1,..., X m be a random vector with the distribution of Xi m 1 N m = n m. By construction, a.s., exactly n m of the variables Xi satisfy Xi > s, and thus m n m satisfy Xi s. Select n + m n m of the latter variables, chosen uniformly at random independent of everything except the set of indices {i : Xi s}, and replace these by variables X i + that are i.i.d. copies of the random variable X + := X 1 X 1 > s and independent of everything else. Denote the result by X m; + then X m + d = X i m 1 N m = n + m. Consequently, by 3.42, P gm X m B m 1, P g m X + m B m

11 Hence, NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 11 P g m X m gm X + m 2B m On the other hand, 3.39 and the construction imply that g m X + m gm X m n + m n m δ > 2K mδ Choosing K = Bδ 1, we obtain a contradiction with Remark 3.5. The constant µ equals E ft by 3.6; we do not know any explicit closed form expression for µ, but it seems possible to use 3.6 for numerical calculation of µ for a given offspring distribution. Note that, by 3.5, ft RT 1, which typically decreases exponentially in the size of T, so convergence ought to be rather fast. For σ 2, [8, 1.17] gives the formula σ 2 = 2 E ft F T T µ Var[fT ] µ 2 / Varξ Again, we do not know any closed form expression, but numerical calculation should be possible. 4. Moments of the number of root subtrees In this section we prove Theorem 2.3, using generating functions and the language of simply generated trees; note that this also shows the equivalent Theorem 2.2. In Sections 4.1 and 4.2, we study a simple example of simply generated trees to illustrate the main idea behind Theorem 2.3; in this example we derive explicit formulas for some generating functions. The proof for the general case is postponed to Section 4.3; it uses the same argument but in general we do not find explicit formulas An example: full binary trees. Consider as an example the simply generated tree T n with the generator Φz := 1 + z 2. Then T n is a uniformly random full binary tree of order n. Provided n is odd; otherwise, such trees do not exist. Note that Φz satisfies the conditions of Theorem 2.3. Note that we have chosen a generator that is not a probability generating function; the corresponding offspring distribution ξ has probability generating function z2, and thus Pξ = 0 = Pξ = 2 = 1 2 ; this generator would lead to similar calculations and the same final result. A combinatorial class is a finite or countably infinite set on which a size function of range Z 0 is defined. For a combinatorial class D and an element δ D, let δ denote its size. The generating function of D is defined by Dz := z δ = d n z n, 4.1 δ D where d n denotes the number of elements in D with size n. It encodes all the information of d n n 0 and is a powerful tool to get asymptotic approximations of d n. Let Z = { } denote the combinatorial class of node, which contains only one element since we are considering unlabelled trees. Let = 1. Then the generating function of Z is simply z. Let F 0 denote the combinatorial class of full binary trees. For T F 0, we let T be the total number of nodes in T. Since T is a binary tree, it must be either a node, or a node n=0

12 12 XING SHI CAI AND SVANTE JANSON together with a left subtree T 1 and a right subtree T 2, with T 1, T 2 F 0. This can be formalized by the symbolic language developed by Flajolet and Sedgewick [6, p. 67] as F 0 = Z + Z F 0 F 0, 4.2 with + denotes or and denotes combined with. Let F 0 z denote the generating function of F 0, i.e., F 0 z := z T = a n z n, 4.3 T n=1 where a n is the number of full binary trees of order n. Then the definition 4.2 directly translates into the functional equation with the explicit solution F 0 z = z + z F 0 z F 0 z = zφf 0 z, 4.4 F 0 z = 1 1 4z z To compute E RT n, we consider a pair T, T in which T is a full binary tree and T is a rooted subtree of T painted with color 1. Let F 1 be the combinatorial class of such partially colored full binary trees, with T, T = T. Let F 1 z be the generating function of F 1, i.e., F 1 z := z T = RT z T =: a 1 n z n. 4.6 T T n=1 T rt Then, for any odd n, E RT n = a 1 n /a n. 4.7 For a tree T in F 1, its root o is always colored. Every subtree T v where v is a child of o so dv = 1 can be either itself a partially colored tree an element of F 1 or an uncolored tree an element of F 0. Thus, we have the following symbolic specification F 1 = Z + Z F 0 + F 1 F 0 + F 1 = ZΦF 0 + F Consequently, using 4.4, F 1 z = zφ F 1 z + F 0 z = z + z F 0 z + F 1 z 2 with the explicit solution = F 0 z + 2zF 0 zf 1 z + zf 1 z F 1 z = 1 1 4zz + F 0 z F 0 z 2z = z 2 1 4z 2 F 0 z z For the second and higher moments we argue similarly. For m 1, we consider a m + 1-tuple T, T 1,, T m in which T is a full binary tree and T 1,, T m are m root subtrees of T with T i painted with color i. Note that T 1,, T m are not necessarily distinct. Note also that a node may have several colors. Let F m be the combinatorial class of such partially

13 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 13 m-colored trees. Let T, T 1,, T m = T. Let F m z be the generating function of F m, i.e., F m z := z T = RT m z T =: a m n z n T T 1,...,T m rt T n=1 Then, for any odd n, E RT n m = a m n /a n The root o of a tree in F m is always painted by all m colors. Every subtree T v where v is a child of o is itself a partially C-colored tree for some set of colors C [m] := {1,..., m}. Let, for a given finite set of colors C, F C be the class of partially C-coloured trees, defined analogously to F m, and note that there is an obvious isomorphism F C = F C. Furthermore, let F m := C [m] F C. Taking into account that there are m k ways to choose k colors out of m, we thus have the equations F m = Z + Z F m F m = ZΦ Fm, 4.13 m m F m = F k. k 4.14 k=0 Consequently, for the corresponding generating functions, F m z = zφ Fm z m 2 m = z + z F k z, 4.15 k which determines every F m z by recursion, solving a quadratic equation in each step. Equivalently, and perhaps more conveniently, m m 1 m m F m z = F k z = F k z + zφ Fm z. k k k=0 m 1 m = 1 m k+1 k k=0 For example, for m = 2, k=0 k=0 F k z + zφ Fm z F 2 z = zφ F 2 z + 2F 1 z + F 0 z = z + z F 0 z + 2F 1 z + F 2 z 2 = z + z F 0 z + 2F 1 z 2 + 2z F0 z + 2F 1 z F 2 z + zf 2 z 2, 4.17 and F 2 z = F 0 z + 2F 1 z + zφ F2 z = F 0 z + 2 F 1 z + z + z F 2 z Explicitly, we obtain from 4.17 or 4.18 F 2 z = z 2z 2 1 4z 2 1 4z z 2 1 4z z 2 1 4z

14 14 XING SHI CAI AND SVANTE JANSON 4.2. Singularity analysis: full binary trees. Let ρ m be the radius of convergence of F m z; then ρ m is a singularity of F m z of square-root type. We see from 4.5 that and thus 1 4ρ 2 0 = 0, 4.20 ρ 0 = Since full binary trees can only have odd number of nodes, we have a 2m = 0 for m 0. For odd n, applying singular analysis to 4.5 gives where λ 0 = 2 π a n = 1 + O n 1 λ 0 n 3 2 ρ n 0, See [6, Theorem VI.2] for details. In fact, in this case we have the well-known exact formula a 2m+1 = C m := 2m!/m! m + 1!, the Catalan numbers [6, p. 67]. Similarly, 4.10 shows that 2 1 4ρ ρ2 1 = 0, 4.23 and thus ρ 1 = = Using the standard singular analysis recipe see [6, Figure VI.7, p. 394], 3+ 3 π a 1 n = 1 + O n 1 λ 1 n 3 2 ρ n 1, where λ 1 = = Such computations can be partially automated with Maple, see, e.g., [9]. Thus 4.7 implies that n ρ E RT n = 1 + O n 1 λ 1 λ 0 For the second moment, 4.19 similarly yields ρ 2 = = Thus a 2 n ρ 1 = 1 + O n 1 λ 2 n 3 2 ρ n 2, 4.28 where λ 2. = is a constant. Then by 4.12 E RT n 2 = 1 + O n 1 λ 2 λ 0 ρ0 ρ 1 n It is not difficult to prove by induction that there exist sequences of numbers λ m > 0 and ρ 0 > ρ 1 > such that for every fixed m 1, and a m n = 1 + O n 1 λ m n 3/2 ρ n m 4.30 E [RT n m ] = 1 + O n 1 λ m λ 0 ρ0 ρ m n. 4.31

15 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 15 This is 2.5 with γ m = λ m /λ 0 and τ m = ρ 0 /ρ m = 2ρ m 1. In particular, τ 1 = = = + 1 = , ρ τ 2 = 1 2ρ 2 = = , 4.33 and γ 1 = = , γ 2 = We do not have a closed form of ρ m or τ m for m 3. Table 1 gives the numerical values of τ m and ρ m for m up to 10. τ τ τ τ τ τ τ τ τ τ ρ ρ ρ ρ ρ ρ ρ ρ ρ ρ Table 1. Numerical values of τ m and ρ m for full binary trees. Remark 4.1. It can be shown, using the equations above and taking resultants to eliminate variables, that ρ 1, ρ 2 and ρ 3 are roots of the equations 16 ρ ρ = 0, ρ ρ ρ ρ = 0, ρ ρ ρ ρ ρ ρ ρ ρ = According to Maple, these polynomials are irreducible over the rationals; moreover, the polynomial in 4.36 is irreducible over Qρ 1 and the polynomial in 4.37 is irreducible over Qρ 1, ρ 2. In particular, we have a strictly increasing sequence of fields Q Qρ 1 Qρ 1, ρ 2 Qρ 1, ρ 2, ρ 3. We expect that this continues for larger m as well, and that the fields Qρ 1,..., ρ m form a strictly increasing sequence for 0 m <. Remark 4.2. The values in show that τ1 2 < τ 2. In fact, τ 2 /τ1 2. = Hence 2.5 implies that, as n, and thus E [ RT n 2] / E [RT n ] Var[RT n ] E [ RT n 2] We expect that the same holds for other conditioned Galton Watson trees, but we have no general proof. Remark 4.3. As said in Remark 2.5, it seems natural to combine Theorems 2.1 and 2.2 and guess that the moments of RT n asymptotically are as the moments of the asymptotic log-normal distribution in Theorem 2.1; this

16 16 XING SHI CAI AND SVANTE JANSON means equality in However, if equality holds in 2.11 for m = 1, 2, 3, then τ1 3 τ2 3 τ 3 = e 3 6+3µ σ2/2 = 1, 4.40 and thus τ 3 = τ2 3 τ Equivalently, ρ 3 = ρ 3 2 ρ 3 1 ρ 0. However, in the case of full binary trees, we have noted in Remark 4.1 that ρ 3 / Qρ 1, ρ 2 = Qρ 0, ρ 1, ρ 2, so 4.41 is impossible. In fact, a numerical calculation, using the values in Table 1, yields in this case τ 3 τ 3 2 τ 3 1 = ρ 1 3 ρ3 2ρ 3 1 ρ 0. = Proof of Theorems Consider a general Φz which satisfies the condition of Theorem 2.3. We define the weighted generating function for m-partially colored trees by F m z := wt z T = wt RT m z T, 4.43 T T 1,...,T m rt T where wt is the weight of T defined in Section 1.3. Note that in case of full binary trees in Section 4.1, wt = 1 and 4.43 agrees with Then we have E RT n m = T : T =n wt RT m T : T =n wt = [zn ]F m z [z n ]F 0 z Following exactly the same argument as in Section 4.1, we have a system of equations m m F m z = zφ F k z, m = 0, 1, k k=0 By induction and the implicit function theorem [6, Theorem B.4], there exist for each m a function F m z that is analytic in some neighborhood of 0 depending on m and satisfies 4.45 there. For singularity analysis, we apply Theorem VII.3 of [6]. We need some preparations. Define again F m z by 4.16, and let and H m z := F m z F m z = m 1 k=0 m k F k z, 4.46 Ψ m z, w := zφ w + H m z Then the implicit equation 4.45 can be written in the equivalent forms F m z = zφ Fm z, 4.48 F m z = Ψ m z, Fm z Let ρ m > 0 be the radius of convergence of F m z, and let s m := F m ρ m. We claim that > ρ 0 > ρ 1 >..., and that for every m, s m < and Ψ m w ρ m, s m =

17 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 17 We prove this claim by induction. The base case m = 0 is well-known, see [6, Theorem VI.6, p. 404], and follows by minor modifications of the argument below. Note first that, by 4.48, F m z R when 0 < z < ρ m, and thus, letting z ρ m, s m + H m ρ m = F m ρ m + H m ρ m = F m ρ m R Next, by 4.47, and, in particular, Ψ m w z, w = zφ w + H m z, 4.52 Ψ m z, Fm z = zφ Fm z w Since F m z has only nonnegative coefficients, it has a singularity at ρ m. This singularity can arise in one of three ways: i ρ m ρ m 1. ii F m ρ m = s m + H m ρ m = R. Recall iii 4.50 holds. In fact, if neither i nor ii holds, then ρ m <, s m < and Ψ m is analytic in a neighbourhood of ρ m, s m. If also iii does not hold, then F m z is analytic in a neighbourhood of ρ m by 4.49 and the implicit function theorem, which contradicts that F m z has a singularity at ρ m. We will show that i and ii are impossible; thus iii is the only possibility. Differentiating 4.49, we obtain F mz = Ψ m z, Fm z + Ψ m z, Fm z F z w mz For 0 < z < ρ m, all terms in 4.54 are positive and finite; hence F mz > z, Fm z F mz and Ψ m w Ψ m z, Fm z < 1, 0 < z < ρ m w Suppose now that i holds. Then F m z is analytic for z < ρ m 1. Furthermore, by induction, F m 1 ρ m 1 = s m 1 <, and H m 1 ρ m 1 <. Hence, F m 1 ρ m 1 = F m 1 ρ m 1 + H m 1 ρ m 1 <. This and the definition 4.16 yield F m ρ m 1 > F m 1 ρ m 1, and thus, using 4.53, Ψ m lim z, Fm z = ρ m 1 Φ Fm ρ m 1 z ρ m 1 w > ρ m 1 Φ Fm 1 ρ m 1 = Ψ m 1 ρm 1, F m 1 ρ m 1 = 1, 4.56 w by the induction hypothesis 4.50 for m 1. However, 4.56 contradicts Hence, i cannot hold, and ρ m < ρ m 1. Next, for 0 < z < ρ m, by 4.48, 4.53 and 4.55, F m zφ F m z Φ F m z = z F m zφ F m z F m z = F m z F m z Ψ m w z, Fm z < F m z F m z. 4.57

18 18 XING SHI CAI AND SVANTE JANSON Since F k z F m z when 0 k m by 4.43, the right-hand side of 4.57 is by 4.16 bounded by 2 m. Suppose now that ii holds. Then, as z ρ m, and thus 4.57 implies ζφ ζ lim ζ R Φζ F m z F m ρ m = R, 4.58 lim sup z ρ m Fm z F m z 2m Consider now two cases. First, if ΦR <, then the left-hand side of 4.59 is RΦ R/ΦR = by the assumption 2.7, which is a contradiction. On the other hand, if ΦR =, then 4.48 and 4.58 yield lim F m z = lim zφ Fm z = ρ m ΦR = z ρ m z ρ m We have shown that ρ m < ρ m 1 ρ k for every k < m, and thus 4.46 shows that H m is analytic at ρ m, and H m ρ m <. Hence, in this case 4.59 yields, using 4.60, ζφ ζ lim ζ R Φζ F m z + H m z H m z lim sup = 1 + lim sup = 1, 4.61 z ρ m F m z z ρ m F m z which contradicts the assumption 2.6. We have thus reached a contradiction in both cases, which shows that ii cannot hold, so F m ρ m = s m + H m ρ m < R Hence, iii holds. Furthermore, by 4.62, s m <, and letting z ρ m in 4.49 yields Ψ m ρ m, s m = s m We now apply [6, Theorem VII.3, p. 468], noting that the conditions are satisfied by the results above, in particular 4.63, 4.50 and This theorem shows that F m z has a square-root singularity at ρ m, and that its coefficients satisfy [z n ]F m z = λ m n 3 ρ n 1 + On 1, 4.64 where λ m > 0 is a constant. In the periodic case, as usual we consider only n such that T n exists. It follows from 4.44 that E RT n m = [zn ]F m z [z n ]F 0 z = λ m m ρ0 1 + O n λ 0 ρ m Letting γ m = λ m /λ 0 and τ m = ρ 0 /ρ m, we have shown 2.5. This prove Theorem 2.3, and thus also the equivalent Theorem A counter example. The following example shows that Theorem 2.3 does not hold without the condition 2.7. Example 4.4. Take the generator m Φz = Φ a z = a + 1 a ζ4 k=1 z k k 4, 4.66

19 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 19 where 0 < a < a 0 := 1 ζ4/ζ3. Then R = 1, ΦR = 1 and zφ z ν := lim z R Φz = Φ 1 = 1 a ζ3 ζ4 = 1 a > 1, a 0 so 2.6 holds. Suppose now that there exists ρ 1 < 1 such that s 1 := F 1 ρ 1 < and Ψ 1 w ρ 1, s 1 = 1, and thus, see 4.53, ρ 1 Φ F 0 ρ 1 + F 1 ρ 1 = Then F 0 ρ 1 + F 1 ρ 1 R = 1. Since F 0 z F 1 z for every z 0, this implies F 0 ρ On the other hand, Φ F 0 ρ 1 +F 1 ρ 1 Φ 1 = ν, and thus 4.68 implies ρ 1 ν 1. Furthermore, F 0 z = zφ F 0 z. Thus, if x := F 0 ρ 1 1 2, we have x = ρ 1 Φx, and thus which yields, recalling 4.67, x = ρ 1 Φx ν 1 Φx, 4.70 Φx = Φ a x νx = 1 a x a 0 We claim that this is impossible if a is close to a 0. In fact, suppose that for every a < a 0 there exists x = x a 1 2 such that 4.71 holds. Then, by compactness, we may take a sequence a n a 0 such that x an converges to some x [0, 1 2 ], and then 4.71 implies Φ a0 x x, 4.72 which is a contradiction since Φ a0 1 = 1 and Φ a 0 x < Φ a 0 1 = 1 for x < x < 1. Consequently, we can find a < a 0 such that the simply generated tree with generator 4.66 does not have F 1 with a singularity of the type above. Hence, in this case, ρ 1 is instead given by ii in Section 4.3, i.e., F 1 ρ 1 = 1, which by 4.48 implies F 1 ρ 1 = ρ 1 Φ F1 ρ 1 = ρ 1, 4.73 F 0 ρ 1 = F 1 ρ 1 F 1 ρ 1 = 1 ρ We have shown that Ψ 1 w ρ1, F 1 ρ 1 < 1, and thus it follows from 4.54 and Φ 1 < that lim z ρ1 F 1 z <. Hence the singularity of F 1 at ρ 1 is not of square root type, and the asymptotic formula 2.5 cannot hold. We leave it as an open problem to find the asymptotics of E RT n and higher moments in this case. 5. General subtrees We have in Sections 4 considered root subtrees. Estimates for general nonfringe subtrees follow from 1.2, but more precise results can be obtained by introducing the corresponding generating functions G m z := wt ST m z T = wt z T, 5.1 T T T 1,...,T m T

20 20 XING SHI CAI AND SVANTE JANSON cf and note that G 0 z = F 0 z. For simplicity, we study first the case m = 1 in detail, and as in Section 4, we consider first the example of full binary trees. We assume throughout this section that the assumptions of Theorem 2.3 hold The mean, full binary trees. Let G 1 be the combinatorial class of pairs of trees T, T such that T is subtree of T F 0. In other words, an element of G 1 is a full binary tree with one non-fringe subtree colored. Such a partially colored tree is either a full binary with a root subtree colored an element of F 1, or a uncolored root together with a left right uncolored subtree an element of F 0 and a partially colored right left subtree an element of G 1. Thus G 1 has the specification G 1 = F 1 + Z F 0 G 1 + Z G 1 F Therefore, G 1 z, the generating function of G 1 given by G 1 z := z T = ST z T =: b 1 n z n, 5.3 T T n=1 T T satisfies G 1 z = F 1 z + 2zF 0 zg 1 z. 5.4 Thus F 1 z G 1 z = 1 2zF 0 z. 5.5 By 5.5, G 1 z has the same radius of convergence ρ 1 as F 1 z, with singularities at the same points ±ρ 1. It is easily verified that the denominator 1 2zF 0 z 0 for z ρ 1, see also Since the singular expansions of the denominator at ±ρ 1 are simply both 1 2ρ 1 F 0 ρ 1, we obtain from 5.5 by singularity analysis b 1 n 1 2ρ 1 F 0 ρ 1 1 a 1 n, 5.6 as n. Thus, using 4.5 and 4.24, a 1 n /b 1 n 1 2ρ 1 F 0 ρ 1 = 3 1. = Therefore the root subtrees form the majority of all subtrees The mean, general trees. For simply generated trees with the generator Φx, 5.2 and 5.4 can be generalized to G 1 = F 1 + Z G 1 Φ F 0, 5.8 and G 1 z = F 1 z + zg 1 zφ F 0 z. 5.9 Therefore, F 1 z G 1 z = 1 zφ F 0 z Note that for any m 0 and 0 < z < ρ m, by 4.53 and 4.50, zφ Fm z < ρ m Φ Fm ρ m = Ψ m w ρ m, s m = Together with F 0 z = F 0 z, this shows that the denominator in 5.10 is non-zero for z < ρ 0, and in particular for z ρ 1. Thus G 1 z has the

21 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 21 same dominant singularities with z = ρ 1 as F 1 z, and it follows that, as n, E RT n E ST n = a1 n 1 ρ 1 Φ F 0 ρ 1 > b 1 n Remark 5.1. In the periodic case, when F 1 z has k 1 singularities on the circle z = ρ 1, it is easily verified that 1 zφ F1 z is a power series in z k and thus has the same value at all these singularities, cf. the full binary case above Higher moments. The generating functions G m for higher moments of ST n can be found recursively by similar methods. The recursion becomes more complicated than for F m, however. We introduce the generating functions for mixed moments of the numbers of root subtrees and general subtrees G m,l := wt ST m RT l z T = wt z T T T T 1,...,T m T, T 1,...,T l rt Note that G m,0 = G m and G 0,l = F l. It follows from 1.2 or the recursions below that G m,l z has the same radius of convergence ρ m+l as F m+l z. Consider first, as examples, the cases with m+l = 2. G 1,1 z is the generating function of the combinatorial class G 1,1 consisting of triples T, T, T where T is a subtree and T a root subtree of T, counted with weights wt determined by Φz by 1.3 and 1.5. Let T, T, T G 1,1. Denote the children of the root o T by v 1,..., v D, and let T 1,..., T D be the corresponding fringe subtrees of T. The subtree T is either a root subtree, and then T, T, T F 2, or it is a subtree of one of the fringe trees T j. Furthermore, the root subtree T is determined by choosing for each fringe subtree T j either a root subtree or nothing. Hence, in the case T, T, T / F 2, for some j 0 D, we choose either T j0, T j 0, T j 0 G 1,1 or T j0, T j 0 G 1,0 ; at the same time, we choose for each j j 0 either T j, T j G 0,1 = F 1 or just T j G 0,0 = F 0. Consequently, and thus G 1,1 = F 2 + Z Consequently, w D D D 1, G 1,1 + G 1,0 F1 + F D=1 G 1,1 z = F 2 z + z G 1,1 z + G 1,0 z Φ F1 z G 1,1 z = F 2z + zg 1,0 zφ F1 z 1 zφ F1 z Similarly, G 2,0 is the class of triples T, T 1, T 2 where both T 1 and T 2 are general subtrees of T. The case when both T 1 and T 2 are root trees gives F 2, and the case where, say, T 1, is a root tree but T 2 is not gives G 1,1 \ F 2, found above. Finally, if neither T 1 nor T 2 is a root tree, then they are

22 22 XING SHI CAI AND SVANTE JANSON determined by one subtree in a fringe tree T j1 and one subtree in T j2, where j 1 and j 2 may be equal or not. This leads to, arguing as in , G 2,0 z = F 2 z + 2 G 1,1 z F 2 z and thus + zg 2,0 zφ F 0 z + zg 1,0 z 2 Φ F 0 z, 5.17 G 2,0 z = 2G 1,1z F 2 z + zg 1,0 z 2 Φ F 0 z 1 zφ F 0 z Singularity analysis of 5.16 and 5.18 show that E[ST n RT n ] E[RT n 2 ] = [zn ]G 1,1 z [z n ]F 2 z 1 1 ρ 2 Φ F 1 ρ 2, 5.19 E[ST n 2 ] E[RT n 2 ] = [zn ]G 2,0 z [z n ]F 2 z 1 1 ρ 2 Φ F 0 ρ ρ 2 Φ F 1 ρ 2 1, 5.20 where all denominators are positive by The argument is easily extended to higher powers, and in principle can any G m,l z be found recursively by this method; however, the formulas will be more and more complicated, and we see no simple general formula. On the other hand, we are really only interested in the singular parts, and thus we can ignore most terms. Lemma 5.2. For each m, l 0, there exist functions φ m,l and ψ m,l that are analytic for z < ρ m+l 1 such that G m,l z = φ m,l zf m+l z + ψ m,l z Furthermore, α m,l := φ m,l ρ m+l satisfies the recursion α 0,l = 1 and, for m 1, α m,l = 1 + ρ m 1 m m+l k=1 k αk,m+l k Φ Fm+l k ρ m+l 1 ρ m+l Φ Fl ρ m+l Proof. The case m = 0 is trivial, with φ 0,l z = 1 and ψ 0,l z = 0. Thus, let m 1. The combinatorial class G m,l consists of sequences T, T 1,..., T m, T 1,..., T, 5.23 where T is a tree, counted with weight wt, each T i is a subtree and each T j is a root subtree. Let k be the number of T 1,..., T m that are not root subtrees. The case k = 0 gives F m+l. Suppose 1 k m. Let again T 1,..., T D be the fringe trees rooted at the children of the root o of T. Further suppose that the k non-root subtrees go into p 1 of T 1,, T D, which we call T j1,, T jp. Then there are D p ways to select Tj1,, T jp. Suppose further that k i of the k non-root subtrees go into T ji. Given p positive integers k 1,, k p with k k p = k, there are k k 1,,k p ways of choosing how the k non-root subtrees are divided among T ji,, T jp. While T jr contains k r marked general subtrees, it also contains i m+l k marked root subtrees, which can be chosen in m+l k i ways. For any Tj such that j / {j 1,, j p }, it too contains up to m + l k marked root l

23 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 23 subtrees. Hence, fixing k, p, k 1,, k p, D, we have the following term that contributes to G m,l z zw D D p k k 1,..., k p p j=1 j=1 m+l k i=0 i=0 m + l k i G kj,iz Fm+l k z D p. Summing over D 1 gives z 1 p m+l k k m + l k G kj,iz Φ p Fm+l k z. p! k 1,..., k p i In 5.25, m + l k m + l 1, and thus Φ p Fm+l k z has radius of convergence at least ρ m+l 1. Similarly, k j + i m + l with equality only if k j = k, and thus j = p = 1, and also i = m+l k; hence, except in the latter case, 5.25 has radius of convergence ρ m+l 1. Consequently, collecting all terms and lumping most of them together, taking into account that there are m k ways to choose the k non-root subtrees among T 1,, T m, G m,l z = F m+l z + z m k=1 m k G k,m+l k zφ Fm+l k z + φz, where φz has radius of convergence ρ m+l 1. Hence, for any m 1, G m,l z = F m+lz + z m 1 m k=1 k Gk,m+l k zφ Fm+l k z + φz 1 zφ Fl z, where the denominator is non-zero for z < ρ m+l 1 ρ l by The results follow by induction in m. Example 5.3. As an example of the recursion 5.26, note that by 5.15 and 5.17, G 2,0 z = F 2 z + 2zG 1,1 zφ F1 z + 2zG 1,0 zφ F1 z + zg 2,0 zφ F 0 z + zg 1,0 z 2 Φ F 0 z, 5.28 where we collect the terms containing G 1,0 z into φz in Proof of Theorem 2.4. An immediate consequence of Lemma 5.2 and singularity analysis, arguing as at the end of the proof of Theorem 2.3 in Section 4.3, cf Note that this yields γ m,l = α m,lγ l+m, 5.29 with α m,l given by the recursion 5.22.

24 24 XING SHI CAI AND SVANTE JANSON 5.4. A numerical example. It follows either from Lemma 5.2 or the direct computations in 5.10, 5.19, 5.20 that 1 α 1,0 = 1 ρ 1 Φ F 0 ρ 1, α 1,1 = 1 ρ 2 Φ F 1 ρ 2, α 2,0 = 1 ρ 2 Φ F 0 ρ 2 1 ρ 2 Φ F 1 ρ Returning again to the full binary trees, we find from by Maple. α 0,1 = ,. α1,1 = ,. α2,0 = , 5.33 γ 0,1. = , γ 1,1. = , γ 2,0. = Hence, recalling Remark 4.2, we find the correlation coefficient between the numbers of root subtrees and general subtrees CovST n, RT n Var STn Var RT n E [ST n RT n ] E [STn 2 ] E [RT n 2 ] γ 1,1 τ 2 n γ 2,0 τ 2 n γ 2τ2 n = γ 1,1 γ 2,0 γ 2 = α 1,1 α2,0. = Therefore, as might be expected, we have a strong but not perfect correlation. Define 6. Average size of root subtrees F 1 z, u := T T rt z T u T =: n=1 a 1 n uz n. 6.1 Then F 1 z = F 1 z, 1 and thus a 1 n = a 1 n 1. Moreover a 1 n u/a 1 n is the probability generating function of X n, where X n = T for a pair T, T where T is a root subtree of T, chosen uniformly at random from all such pairs in F 1 with T = n. F 1 z, u can be computed in the same way as in Section 4. Example 6.1. In the case of full binary trees, we obtain F 1 z, u = zuφ F 1 z, u + F 0 z = zu + zu F 0 z + F 1 z, u 2 = uf 0 z + 2zuF 0 zf 1 z, u + zuf 1 z, u 2, 6.2 with the explicit solution F 1 z, u = 1 1 4zuzu + F 0 z F 0 z 2zu = 1 2u 1 4z u 4z 2 u 2 F 0 z zu

25 NON-FRINGE SUBTREES IN CONDITIONED GALTON WATSON TREES 25 By the well-known relationship between derivatives of probability generating functions and factorial moments [6, p. 158], we have E X n r = dr a 1 n u du r = [zn ] r F 1 z, u/ u r u=1 [z n, 6.4 ]F 1 z u=1 a 1 n where x r := xx 1 x r + 1. It is not difficult to use induction and singularity analysis of the partial derivatives r F 1 z, u/ u r u=1 to show that [z n ] r F z, u u r = 1 + O n 1 2 r λ 1 n r 3 2 ρ n 1 u=1 3, 6.5 where λ 1 is as in Thus it follows from the estimation of a 1 n in 4.25 that EX n r = 1 + O n 1 2n r Computing one more terms for E [X n 2 ] yields, omitting the calculations, Var X n n =: σ 2 9 1n. 6.7 The asymptotic estimates 6.6 and 6.7 suggest, but are not enough to conclude, asymptotic normality. However, we can apply the following general theorem. Theorem 6.2. Assume Φz satisfies the conditions of Theorem 2.3. Let ρ 1, s 1 and F 0 z be as in Section 4.3. Let and µ 1 := d 1 := F 0ρ 1, d 2 := F 0 ρ 1, d 3 := Φ s 1 + F 0 ρ 1, 6.8 s 1, σ1 2 s 1 + ρ 1 d := ρ 1 d2 d 3 ρ 1 s d2 1 d 3ρ 1 s 1 + d 1 d 3 s d2 1 d 1 ρ 1 + s d 3 Then 0 < µ 1 < 1 and Furthermore, as n, E [X n ] = µ 1 n + O1, Var X n = σ 2 1n + O X n µ 1 n n d N0, σ 2 1, 6.11 and if σ 2 1 0, then also X n E [X n ] Var Xn d N0, Proof. Let F x, y, u := xuφy + F 0 x. Then most conditions of [5, Theorem 2.23] are satsifed by the arguments in Section 4.3, in particular We only need to verify that F x ρ 1, s 1, 1 0, 2 F y 2 ρ 1, s 1, 1 0, 6.13 where ρ 1 and s 1 are as in Section 4.3. The first inequality holds since F x ρ 1, s 1, 1 Φ s 1 + F 0 ρ 1 = s 1 > ρ 1