You are on page 1of 53

COUNTING CLOSED GEODESICS IN STRATA

ALEX ESKIN, MARYAM MIRZAKHANI, AND KASRA RAFI
arXiv:1206.5574v1 [math.GT] 25 Jun 2012

Abstract. We compute the asymptotic growth rate of the number N (C, R)
of closed geodesics of length ≤ R in a connected component C of a stratum of
quadratic differentials. We prove that for any 0 ≤ θ ≤ 1, the number of closed
geodesics γ of length at most R such that γ spends at least θ–fraction of its
time outside of a compact subset of C is exponentially smaller than N (C, R).
The theorem follows from a lattice counting statement. For points x, y in the
moduli space M(S) of Riemann surfaces, and for 0 ≤ θ ≤ 1 we find an upper-
bound for the number of geodesic paths of length ≤ R in C which connect a
point near x to a point near y and spend a θ–fraction of the time outside of a
compact subset of C.

1. Introduction
Let S = Sg,p be a surface of genus g with p punctures and let M(S) be the moduli
space of Riemann surfaces homeomorphic to S. The co-tangent bundle of M(S) is
naturally identified with QM(S) the space of finite area quadratic differential on S.
Let Q1 M(S) be subspace of quadratic differentials with area 1. There is a natural
et

0
SL(2, R) action on the Q1 M(S). The orbits of the diagonal flow, gt = ,
0 e−t
projects to geodesics in M(S) equipped with the Teichmüller metric. For R > 0,
let N (R) be the number of closed Teichmüller geodesics of length less than or equal
to R on Q1 M(S). It was shown in [EM2] that, as R → ∞ the number N (R) is
asymptotic to ehR /hR, where h = 6g − 6 + 2p.
The moduli space of quadratic differentials is naturally stratified: to each qua-
dratic differential (x, q) ∈ QM(S) we can associate σ(q) = (νi , . . . , νk , ς) where
ν1 , . . . , νk are the orders of the zeros and poles of q, and ς ∈ {−1, 1} is equal to 1
if q is the square of an abelian differential and −1 otherwise. For a given tuple σ,
we say a quadratic differential (x, q) ∈ QM(S) is of type σ if σ(q) = σ. The space
QM(σ) of all quadratic differentials in QM(S) of type σ is called the stratum of
quadratic differentials of type σ. The stratum QM(σ) is an analytic orbifold of
real dimension 4g + 2k + ς − 3.
Let Q1 M(σ) be the space of quadratic differentials in QM(σ) of area 1. It is
not necessarily connected (see [KZ] and [La] for the classification of the connected
components), however, each connected component is SL(2, R) invariant. Let C be a
connected component of Q1 M(σ). In this paper, we study the asymptotic growth
rate of the number N (C, R) of closed Teichmüller geodesics of length less than
or equal to R in C. Our main tool is estimating the number N (CK , R) of closed
geodesics that stay completely outside of a large compact set K ⊂ C.
Theorem 1.1. Given δ > 0 there exists a compact subset K ⊂ C and R0 > 0 such
that for all R > R0 ,
N (CK , R) ≤ e(h−1+δ)R .
1

COUNTING CLOSED GEODESICS IN STRATA 2

This result implies that:
Theorem 1.2. As R → ∞, we have
ehR
N (C, R) ∼ ,
hR
where h = 12 [1 + dimR (C)] and the notation A ∼ B means that the ratio A/B tends
to 1 as R tends to infinity.
Remark 1.3. In case of abelian differentials, h is equal to the dimension of the
relative homology of S with respect to the set of singular points of (x, q) ∈ C,
otherwise h is one less.
Recurrence of geodesics. We prove a stronger version of Theorem 1.1. Every
quadratic differential defines a singular Euclidean metric on the surface S and for
every compact set K ⊂ C, there is a lower bound for the q–length of a saddle
connection where q ∈ K. Here, we restrict attention to closed geodesics where more
than one simple closed curve or saddle connection is assumed to be short; in this
case the growth rate is of even lower order.
Let T (S) be the Teichmüller space, the universal cover of M(S). Let QT (S)
and Q1 T (S) be defined similarly. To distinguish between point in the Moduli
space and Teichmüller space, we use x ∈ M(S) and X ∈ T (S). Also, we use
(x, q) ∈ Q1 M(S), (X, q) ∈ Q1 T (S) denote the geodesics in Q1 M(S) by g and
the geodesics in Q1 T (S) by G. The space Q1 T (S) is also naturally stratified. We
denote the space of of quadratic differentials in Q1 T (S) of type σ by Q1 T (σ). To
simplify the notation, let
Q(σ) := Q1 T (σ).
Recall that ExtX (α) denotes the extremal length of a a simple closed curve α on
the Riemann surface X ∈ T (S). (see Equation (1) for definition). We introduce
a notion of extremal length for saddle connections on quadratic differentials. Es-
sentially, the extremal length of a saddle connection ω in a quadratic differential
(X, q) ∈ Q1 T (S) with distinct end points p1 and p2 is the extremal length of the
associated curve in the ramified double cover of X with simple ramification points
at only p1 and p2 (see §4.5 for more details).
.
Definition 1.4. For  > 0 and for any quadratic differential (X, q) ∈ Q(σ), let
Ωq () be the set of saddle connections ω so that either Extq (ω) ≤  or ω appears
in a geodesic representative of a simple closed curve α with ExtX (α) ≤ . Let
Qj, (σ) be the set of quadratic differentials (X, q) of type σ so that Ωq () contains
at least j disjoint homologically independent saddle connections. When σ is fixed,
we denote this set simply by Qj, . Let Cj, be the subset of C of unmarked quadratic
differentials whose lift to Q1 T (S) is lies in Qj, . For 0 ≤ θ ≤ 1, define Nθ (Cj, , R)
to be the number of closed geodesics of length at most R in C that spend at least
θ–fraction of time in Cj, .
In this paper, we show:
Theorem 1.5. Given δ > 0 there exists  > 0 small enough and R > 0 large
enough so that for all j ≥ 1 and 0 ≤ θ ≤ 1,
Nθ (Cj, , R) ≤ e(h−jθ+δ)R .

COUNTING CLOSED GEODESICS IN STRATA 3

Remark 1.6. The condition on Ωq () is necessary. Just assuming there are j saddle
connections of q–length less than  does not reduce the exponent by j. In fact,
for any , there is a closed geodesic g in Q1 M(S) where the number of saddle
connections with q–length less than  is as large as desired at every quadratic
differential (X, q) along g. This is because the Euclidean size of a subsurface could
be as small as desired (see §4.4) and short saddle connection can intersect.
Lattice counting in Teichmüller space. Let Γ(S) denote the mapping class
group of S and let BR (X) denote the ball of radius R in the Teichmüller space with
respect to the Teichmüller metric, centered at the point X ∈ T (S). It is known
([ABEM]) that, for and Y ∈ T (S),

Γ(S) · Y ∩ BR (X)

∼ Λ2 e(6g−6)R ,

.

Here Λ is a constant depending only on the topology of S (See [Du]). Here we are interested in the case where the Teich- müller geodesic joining X to g·Y .4). X. X. there exists  > 0 small enough and R > 0 large enough such that for every 0 ≤ θ ≤ 1. the restriction of the Teichmüller flow to any connected component C of Q1 M(σ) is mixing with respect to the Lebesgue measure class [Ma]. for a fix r0 > 0. R) be the number of points Z ∈ T (S) such that: • Z ∈ BR (X) and Z = g · Y . as R → ∞. However. Similar method has been used in [EM2] where it is enough to use Minsky’s product region theorem (see §2. we define SX to be the set of 0 –short curves in X and Y 1 G(X) = 1 + p . Y ∈ T (S) we have Nθ (Qj. for some g ∈ Γ(S). • there is a Teichmüller geodesic segment G ⊂ Q(σ) joining X1 ∈ Br0 (X) to Y1 ∈ Br0 (Y ) • G spends at least θ–fraction of the time in Qj. Notes on the proof. More precisely. Also. invariant by the Teich- müller flow which is equivalent to Lebesgue measure. Y.1) in this paper. the Teichmüller flow on C is exponentially mixing with respect to µ [AR]. The main difficulty for proving Theorem 1.2 in [EM2]. Compare with Theorem 7.2 is the fact that the Teichmüller flow is not uniformly hyperbolic. for g ∈ Γ(S). is assumed to belong to the stratum Q1 M(σ) or stay close to it. . Each stratum Q1 M(σ) has an affine integral structure. The main theorem in this paper is a partial generalization of this result for the strata of quadratic differentials. 1. Given δ > 0. we will only use the mixing property (as stated in Theorem 3. Moreover. since the projection of C .3) to prove the necessary estimates. . called the Masur-Veech measure. for a fix 0 . X. Y. [AGY]. β∈SX ExtX (β) Theorem 1. In this paper. . and carries a unique probability measure µ. 2. [Ve]. In fact.7. let Nθ (Qj. R) ≤ G(X)G(Y ) e(h−jθ+δ)R . As in [EM2] we show that the Teichmüller geodesic flow (or more precisely an associated random walk) is biased toward the part of C that does not contain short saddle connections (see Lemma 7. j ≥ 1.

near points with degenerating zeros of the quadratic differential.7 and the order in which the constants in the argument are chosen. We choose our constants as follows: We call a curve short if its extremal length is less than 0 . qa ) and (Xb . Such a triangulation captures the geometry of singular Euclidean metric associated to q in a way that is compatible with the hyperbolic metric associated to X. Given λi . It turns out that the number of possible triangulations Tb is related to the number of edges in Ta that are homologically independent. τ )– regular triangulation for a quadratic differential (X. 3. . The distance between two points in the trajectory is at most τ . Theorems 1. We remark that the Hodge norm behaves badly near smaller strata. τ )–regular triangulation and in §5 we establish the necessary bounds on the intersection number between Ta and Tb needed in §6.13 holds and  ≤ 2 = 2 (τ ) so that Lemma 7. 5.5 are essentially corollaries of Theorem 1. The dependence on the choice of τ and  is always highlighted and a constant that we call universal does not depend on  or τ . We note that Nθ (Qj. τ )–regular triangulation Tb so that Ta and Tb have j-edges in common. q) (Definition 4. for the geodesic segment G : [a. with initial and terminal quadratic differentials (Xa . In fact. R) is less than then number of trajectories {λ0 . i. . one can find a (qa .2. COUNTING CLOSED GEODESICS IN STRATA 4 in M(S) is not easy to understand. we fix a net N in M(S) and its lift Ñ in T (S).4 in §7) Then  is chosen small enough depending on the value of τ . the condition that the segment [λi . This is the main reason that the growth rate of Nθ (Qj. we bound the number of possible choices for λi+1 . . Theorem 6. λi+1 ] can be approximated by a path in Qj.7. where quadratic differentials have small geodesic segments. For the arguments in §7 to work. 4. In §7. In order to prove Theorem 6. . will also reduce the number of choices for λi+1 in the following way. We define a notion of a (q. we use the basic properties of the Hodge norm [ABEM] to prove a closing lemma for the Teichmüller geodesic flow in §9. To finish the proof of Theorem 1.1 holds. In §4 we establish the basic properties of a (q. τ )–regular triangulation (Lemma 4. We need  ≤ 1 = 1 (τ ) so that Lemma 4. X. .2 and the estimate in Equation (5) hold. we need to choose τ large enough depending on the value of δ (See proofs of Theorem 1. we use Theorem 6. Most of these constants are hidden in notations  and ≺ (see the notation section below). λn } in Ñ from X to Y so that for θ proportion of steps the segment [λi . given the triangulation Ta . the number of possible triangulations Tb which have a certain bounds on their intersection number with Ta . X. τ )–regular triangulation Ta and (qb .1 (§6) we compute.e.13). R) is related to dimR C.1 and 1. We call any other constant that depends in the topology of S or the choice of 0 ∗ ∗ a universal constant.1 to prove Theorem 1. We will show that a set of disjoint saddle connections in Ωq () can be included in a (q.11). But. we need different and more delicate methods to obtain similar results. b] → Qj. First.1 shows that this condition reduces the number of choices for λi+1 by a factor e−jτ . . qb ).5 and Lemma 7. Y. The bound depends on the geometry of λi (captured by the function G(  ). . .7. This is a constant that depends on the topology of S only (a uniform constant) and is chosen so that Theorem 2. λi+1 ] can be approximated by a path in Qj. Here we describe the steps involved in the proof of Theorem 1. Y.

By a theorem of Bers. the number of closed geodesics γ of length at most R such that γ spends at least a θ–fraction of the time outside of a compact subset of C is exponentially smaller than N (C. 3. Hence the number of conjugacy classes of pseudo- Anosov elements of the group Γ(S) with dilatation factor Kg ≤ eR is finite. the geodesic flow is uniformly hyperbolic (See [Ve]. . R) is the number of conjugacy classes of pseudo-Anosov elements g in the mapping class group Γ(S) with expansion factor of at most eR such that Gg ⊂ C.  L(S) = log(Kg ) | g ∈ Γ(S) pseudo-Anosov is the length spectrum of M(S) equipped with the Teichmüller metric. We remark that for any pseudo-Anosov g ∈ Γ(S) the number Kg is an algebraic number and log(Kg ) is equal to the minimal topological entropy of any element in the same homotopy class [FLP]. Foliations fixed by pseudo-Anosov maps can be characterized by being repre- sentable by eventually periodic "convergent" words [PP1]. [Fo] and §8). [Ath]. there is an inequality relating the length of the repeating part of the word corresponding to a pseudo-Anosov foliation and the dilatation factor of a pseudo-Anosov map preserv- ing that foliation [PP2]. Variations on this theme have been used in [EMM]. N (C. Let Kg denote the dilatation factor of g [Th]. and [EM2]. in view of Theorem 1. for any 0 ≤ θ ≤ 1. COUNTING CLOSED GEODESICS IN STRATA 5 On the other hand. every irreducible mapping class element g ∈ Γ(S) of infinite order has a representative which is a pseudo-Anosov homeomorphism. "most" closed geodesics spend at least (1 − θ)–fraction of the time away from the degenerating locus. In this paper. The first results on this problem are due to Veech [Ve]. One reason the proof is different from [EM2] is that in general the projection map π : Q1 M(σ) → M(S) is far from being a fibration: in many cases dim(Q1 M(σ)) < dim(M(S)) and dim(π −1 (X)∩Q1 M(σ)) depends on the geometry of X. every closed geodesic in M(S) is the unique loop of minimal length in its homotopy class. The basic idea behind the proof of the main theorem in this paper is proving recurrence results for Teichmüller geodesics. By [AY] and [Iv]. Also a pseudo-Anosov g ∈ Γ(S) gives rise to a closed geodesic Gg of length log(Kg ) in Q1 M(S). 2. the estimates obtained using these inequalities are weaker. R). 1. According to the Nielsen-Thurston classification. Moreover. Hence log(Kg ) is the translation length of g as an isometry of T (S) [Be].2 following the ideas from Margulis’ thesis [Mar].5. However. This allows us to prove Theorem 1. The results obtained in §4 allow us to deal with this issue. In other words. L(S) is a discrete subset of R. Further remarks and references.) In terms of this notation. He proved that there exists a constant c such that log N (R) log N (R) h ≤ lim inf ≤ lim sup ≤c R→∞ R R→∞ R and conjectured that c = h. the set of quadratic differentials with no small geodesic segment is compact and in any compact subset of C. Also. (See [Pe] and [BC] for simple explicit constructions of pseudo-Anosov mapping classes. Therefore. [EM1]. we need to analyze the geometry of quadratic differentials more carefully.

A  B if + + ∗ both A ≺ C and B ≺ A hold. ∗ + Notation. A point in the Teichmüller space T (S) is a Riemann surface X of genus g with p marked points equipped with a diffeomorphism f : S → X sending marked points to marked points. COUNTING CLOSED GEODESICS IN STRATA 6 4. R) ≥ e(h−1)R . Our results are complimentary to the following result: Theorem 1. The mapping class group Γ(S) acts on T (S) by changing the marking. The quotient space M(S) = T (S)/Γ(S) is the moduli space of Riemann surfaces homeomorphic to S. Rafi). see [Hu] and [FM]. the space n . The notation A = O(B) means that A ≺ B.1. the expression A ≺ B means that A < c B and A ≺ B means A ≤ B + c for some universal constant c which only depends on the topology ∗ ∗ ∗ + of S. Similarly. 2 f where f : X1 → X2 ranges over all quasiconformal maps isotopic to f1 ◦ f2−1 and Kf ≥ 1 is the dilatation of f . we will often omit the marking and write X ∈ T (S). by results in [H3] the normalized geodesic flow invariant measure supported on the set of closed geodesics of length ≤ R in C become equidistributetd with respect to the Lebesgue measure µ as R → ∞. 2. diffeomorphic to a cell. Teichmüller metric and extremal lengths In this section. Two marked surfaces f1 : S → X and f2 : S → Y define the same point in T (S) if and only if f1 ◦f2−1 : Y → X is isotopic (relative to the marked points) to a holomorphic map. N (CK . Also. For convenience. The map f provides a marking on X by S. we use small case letters for points in Moduli space and write x ∈ M(S). By the uniformization theorem. We write A  B if we have both A ≺ B and B ≺ A. To distinguish between a marked point and an un-marked point. 2. the group of isotopy classes of orientation preserving self-homeomorphisms of S fixing the marked points pointwise. each point X in T (S) has a complete metric of constant curvature −1 with punctures at the marked points. The space T (S) is a complex manifold of dimension 3g − 3 + p. Let Γ(S) denote the mapping class group of S.8 (Hammenstadt. For more details. Teichmüller space. (f2 : S → X2 ) = inf log(Kf ). In this paper. There exists a compact K ⊂ C such that for R sufficiently large. we recall some definitions and known results about the geometry of M(S) equipped with the Teichmüller metric. The Teichmüller met- ric on T (S) is defined by  1 dT (f1 : S → X1 ). The cotangent space of T (S) at a point X can be identified with the vector space Q(X) of quadratic differentials on X with at most simple poles at the marked points of X. Let S be a connected oriented surface of genus g with p marked points.

o QT (S) = (X. q) .

q ∈ Q(X) . X ∈ T (S).

q is the tensor given by q(z)dz 2 . is the cotangent space of T (S). where q(z) is a meromorphic function with poles of degree at most one . In local coordinates z.

β) ≤ ExtX (α) · ExtX (β). In the interest of brevity. Given curves α and β on S. ρ) where the supremum is taken over all metrics ρ conformally equivalent to X. Similarly. To simplify the notation. in this paper. and p : Q1 T (S) → Q1 M(S). π will denote both projection maps: π : Q1 M(S) → M(S). β∈S ExtY (β) The relationship between the extremal length and the hyperbolic length is com- plicated. COUNTING CLOSED GEODESICS IN STRATA 7 at the punctures of X. Extremal and hyperbolic lengths of simple closed curves. we refer to α simply as a curve on S. in general. or equivalently the unit cotangent bundle over T (S). Define QM(S) ∼ = QT (S)/Γ(S) and Q1 M(S) ∼ = Q1 T (S)/Γ(S). and π : Q1 T (S) → T (S). we let p denote both projection maps p : T (S) → M(S). the Teichmüller metric corresponds to the norm Z k q kT = |q(z)| |dz|2 X on QT (S). In general. `X (α)2 ≤ ExtX (α). and `ρ (α) denotes the infimum of ρ–length of a representative α. by the definition of extremal length. β) is the minimum number of points in which representatives of α and β must intersect. Given X. the Teichmüller distance between X and Y is given by p ! ExtX (β) dT (X. the intersection number i(α. non-peripheral. The extremal length of a curve α on X is defined by `ρ (α)2 (1) ExtX (α) := sup . 2. Y ∈ T (S). Y ) = sup log p . let `X (α) denote the hyperbolic length of the unique geodesic in the homotopy class of α on X. The homotopy is always relative to the marked points. Let Q1 T (S) denote the space of (marked) unit area quadratic differen- tials. We denote the set of curves on S by S to emphasize that they are simple curves. simple closed curves on the surface S. by [GM] p p (2) i(α.1 (Kerckhoff). Given a curve α on the surface S and X ∈ T (S). The following result [Ker] relates the ratios of extremal lengths to the Teich- müller distance: Theorem 2.2. ρ Area(X. even an open annulus. π(2g − 2 + p) . ExtA (α) where α is the core curve of A. Recall that the modulus of an annulus A is defined to 1 Mod(A) := . Let α rep- resent the free homotopy class of non-trivial. In this setting. Here X can be any Riemann surface.

As a result. for any X ∈ T (S). Minsky’s product theorem. the extremal length can be extended continuously to the space of measured laminations [Ker] such that ExtX (r · λ) = r2 ExtX (λ). since the space of projectivized measured laminations is compact.3. ExtX (α) 2. π `X (α) 2 Hence. `X (α) ∗  1. for a fixed 0 . for every X there exists a constant cX so that 1 p `X (α) ≤ ExtX (α) ≤ cX `X (α). as `X (α) → 0. cX However. . by [Mas] 1 ExtX (α) 1 (3) ≤ ≤ e`X (α)/2 . . COUNTING CLOSED GEODESICS IN STRATA 8 Also. . . Let A = {α1 . αj } be a collection of disjoint simple closed curves on S and. n .

o T0 (A) = X ∈ T (S) .

1 ≤ i ≤ j . . ExtX (αi ) ≤ 0 .

using the Fenchel-Nielsen coordinates on T (S). More precisely.  ) denote the supremum metric on (H2 )j × T (S \ A). Then: Theorem 2. . There is 0 > 0 is small enough and B > 0 depending only on S such that for all X. `X (α1 ) `X (αj ) Here. Y ∈ T0 (A). Following Minsky.2 (Minsky). The product region theorem [Mi] states that for sufficiently small 0 the Teich- müller metric on T0 (A) is within an additive constant of the supremum metric on (H2 )j × T (S \ A). Then. θj (X). we can define ΦA : T0 (A) → (H2 )j by   1 1 ΦA (X) = θ1 (X). . let dA (  . . . . where T (S \ A) is the quotient Teichmüller space obtained by collapsing all the αi . we get a map Φ : T0 (A) → (H2 )j × T (S \ A). . . θi (  ) is the Fenchel-Nielson twist coordinate around αi and represents the x–coordinate in upper-half plane H and the y–coordinate in H is the reciprocal of the hyperbolic length.

.

.

Y ) − dA Φ(X). Φ(Y ) .dT (X.

As mentioned in the introduction. . < B. we fix 0 so that the above theorem and the estimate in Equation (5) hold.

Lemma 2. if two curves are short in X they can not intersect. Let SX be the set of short curves on X. From discussions in [Mi]. COUNTING CLOSED GEODESICS IN STRATA 9 2. We also recall the following lemma which.3. Y ) = O(1) then G(X)  G(Y ). we know that. follows from [EM2].4. For any X ∈ T (S) let  . for example. Short curves on a surface. For 0 as above. The function G is Γ(S) invariant and induces a proper function on M(S). Define G : T (S) → R+ by Y 1 (4) G(X) = 1 + p . we say a curve α is short on X if ExtX (α) ≤ 0 . α∈SX ExtX (α) ∗ If dT (X.

IX = g ∈ Γ(S) .

Then . be the set of mapping classes that move X by a bounded amount. dT (g · X. X) = O(1) .

.

∗ .

IX .

however. The homotopy class of an arc (relative to its endpoints) has a unique q–geodesic representative. see [St]). Moduli space of quadratic differentials and Teichmüller’s theorem 3. 3. Moduli space of quadratic differentials. 3. . In this case. q) 1 be a quadratic differential. . . νk ). νk . q) is a q–geodesic segment which connects a pair of singular points without passing through one in its interior. q) of quadratic differentials on X with zeros and poles of multiplicities (ν1 . ς = −1. . We denote the Euclidean length of a saddle connection ω on q by `q (ω). 1} is equal to 1 if q is the square of an abelian differential (an abelian differential). If we represent q locally as q(z)dz 2 then |q| = |q(z)| 2 |dz| defines a singular Euclidean metric on X with cone points at zeros and poles. . As a result. The total angle at a singular point of degree ν is (2+ν)π. Then G QT (S) = QT (σ).  G(X)2 . (for more details.1. Otherwise. For σ = (ν1 . ς) define QT (σ) ⊂ QT (S) to be the subset consisting of pairs (X. Flat lengths of simple closed curves and saddle connections. The sign ς ∈ {−1. See [Ma] and [MS2] for more details. .2. . We refers to the geodesic representative of a curve or an arc in this singular flat metric as the q–representative of that curve or arc. Any curve α either has a unique q–geodesic representative or there is flat cylinder of parallel representatives. Although the value of q ∈ Q(X) at a point in X depends on the local coordinates. Let (X. we say α is a cylinder curve and we denote the cylinder of geodesics representatives of α by Fα . . A saddle connection on (X. there is a natural stratification of the space QT (S) by the multiplicities of zeros of q. The poles are always assumed to be simple and are located at the marked points. the zero set of q is well defined. not all marked points have to be poles. σ It is known that each QT (σ) is an orbifold. We denote the Euclidean length of the q–representative of α by `q (α). .

Σ. A · q ∈ Q1 T (S) is determined by the new atlas {Aφi }. pk } to R2 such that the change of coordinates are of the form v → ±v + c. we have ([Ve]. R)) − 1 independent vectors of E. R) acts naturally on Q1 T (S) by acting on the corresponding atlas. In case of abelian differentials (ς = 1) it is enough to choose a basis for H1 (S. R) if ς = 1. any saddle connection ω joining two zeros of a quadratic differential √ q = ζdz 2 determines a complex number holq (ω) (after choosing a branch of ζ and an orientation of ω) by     Z p Z p holq (ω) =  < ζ  +  Im ζ  i. We denote both actions (on Q1 T (S) and Q1 M(S)) by gt . 3. . Each connected component C of a stratum Q1 M(σ) carries a unique probability measure µ in the Lebesgue measure class such that: . Moreover.3. Σ) with |Σ| = k. . Note that for non-orientable differentials (ς = −1) there will be a linear relation between the holonomies of the vectors corresponding to a basis for the relative homology (see §5.4.q is linear.3). R) of the pair (S. However. νk . and  h = (2g + k − 2) = dim H1 (S. One can choose h directed edges ω1 . Σ. R). Σ. In this case. .   This action descends to Q1 M(S) via the projection map p : Q1 T (S) → Q1 M(S). Σ. Σ.q ◦ ψ −1 T.1 (Veech-Masur).   and = holgt q (ωi ) = e−t = holq (ωi ) .q (q) = holq (ωi ) i=1 is a local homeomorphism. Consider the first relative homology group H1 (S. . The et  0 action of the diagonal subgroup gt = is the Teichmüller geodesic flow 0 e−t for the Teichmüller metric. For a more detailed discussion of the holonomy coordinates see [MS1]. We recall that given (X. q) ∈ Q1 T (σ) there is a triangulation T of the underlying surface by saddle connections. . it is enough to choose dim(H1 (S. . in holonomy coordinates the Teich- müller flow is simply defined by < holgt (q) (ωi ) = et < holq (ωi ) . . Let  h = (2g + k − 1) = dim H1 (S. Teichmüller geodesic flow. COUNTING CLOSED GEODESICS IN STRATA 10 3. given A ∈ SL(2. Period coordinates on the strata. . ωh of T . R) from the edges of E. . . . We recall that when 3g + p > 4 the Teich- müller metric is not even Riemannian. Also for any other geodesic triangulation T 0 the map the map ψ T 0 . and an open neighborhood Uq ⊂ QT (σ) of q such that the map h ψ T. R) − 1 if ς = −1. . q) ∈ Q1 T (S) with zeros at p1 . ς) the period coordinates gives QT (σ) the structure of an affine manifold. ω ω We recall that for any σ = (νi . The subspaces Q1 T (σ) and Q1 M(σ) are invariant under the Teichmüller geodesic flow. pk is determined by an atlas of charts mapping open subsets of S − {p1 . [Ma]): Theorem 3. In general. . Therefore the group SL(2. In other words. . geodesics in this metric are well understood.q : QT (σ) → Ch Defined by ψ T. . A quadratic differential (X.

α and β may share an edge. R) is volume preserving and ergodic. 4. In the hyperbolic metric of X.13 which shows the existence of such triangula- tions. one can naturally move from the intersection number between homotopy classes of curves to inter- section number between their geodesic representatives. Moreover. we establish some basic properties of (q. we always mean that they have an essential intersection not tangential. One can give an estimate for the extremal length of a simple closed curve α in X by ex- amining the singular Euclidean metric |q|. q) using the saddle connections that captures the geometry of the singular Euclidean metric associated to q. this is not true. X2 ). there exist quadratic differentials (Xi . The intersection number between a saddle connection and itself is zero. 3. gd (q1 ) = q2 . The main statement of the section is Lemma 4. τ )–regular triangulations. Given any X1 . To be more precise. If A is an annulus. Teichmüller’s theorem. when we say α and β intersect. (saddle connections and curves) the intersection number is denoted by i(  .1. To intersect A. The curve α crosses A if α enters one side of A and exits the other. which can be calculated from the conformal structure of X. in the annular cover X̃A of X associated to α. • Teichmüller geodesic flow is mixing. we distinguish between a curve α intersecting A and crossing it. Hence. See [FM] and [Hu] for more details. τ )–regular triangu- lation. we say two saddle connections are disjoint if they have disjoint interiors or they are equal. X2 ∈ T (S). α needs only to enter the interior of A. 4. d = dT (X1 . In both cases. Intersection number. We also define the notion of a (q. To simplify the exposition. and the map f is the same as the map induced by the Teichmüller flow gd . We recall the following important theorem due to Teichmüller. This is a partial triangulation of (X. we mean the set of points that are q–distance at most c from this set for some constant . the geodesic represen- tative of any two curves α and β intersect minimally. Here.5. we mean that any lifts α̃ and β̃ to the universal cover q̃ of q have end points in the boundary that do no interlock.2. Geometry of a quadratic differential In this section. By a regular neighborhood of a set. X2 ) = 12 log(Kf ). the quadratic differential in X1 in the direction of X2 . qi ) ∈ Q1 (Xi ) such that the map f takes zeroes and poles of q1 to zeroes and poles of q2 of the same order. there is a lift of α connecting the two boundary components of X̃A . That is. there exists a unique quasi-conformal map f . In the singular Euclidean metric.  ). We denote q1 by q(X1 . We describe how the extremal length of a curve. called the Teichmüller map. 4. relates to the singular Euclidean metric associated to (X. Extremal lengths and flat lengths of simple closed curves. In other words. In the rest of the section. q). We also talk about the intersection number between two saddle connections. However. The intersection number between two saddle connection is the number of interior intersection points. COUNTING CLOSED GEODESICS IN STRATA 11 • the action of SL(2. these intersections are tangential. the q–geodesic representatives of curves α and β that have geometric intersection number zero may intersect. By this. But they do not cross. q). we recall some of the basic geometric properties of a quadratic differential (X.

We need to measure what is the largest neighborhood of α that still has a simple topology. We intend Log to apply only to large numbers.2. We make this precise: Definition 4. There is an implicit assumption that τ is large enough (say. we say a curve α is short in q if ExtX (α) ≤ 0 . is the distance between steps of a random walk trajectory. . In what follows. Short simple closed curves. Denote the set of large-cylinder curves by Sq≥τ and define Sq≤τ = Sq \ Sq≥τ . Fα and Gα respectively. We say α is a cylinder curve if the interior of Fα is not empty. Let τ be a positive real number and let Mτ = e−2τ .3. the size sα of Fα is defined to be equal to the distance between the boundaries of Fα . 4. (In a degenerate case. the length of a curve α ∈ Sq≤τ changes slowly while the modulus of Fα remains small. the value of either e. We use the modified logarithm to avoid this issue. a simple closed curve that has a short flat length may not have a small extremal length. That is. α behaves essentially like a non-cylinder curve. We say a curve α ∈ Sq is a large-cylinder curve if Mod(Fα ) ≥ Mτ . ≥τ For α ∈ Sq .13 and Equation (5) hold) we have the following estimate (see [R3]) 1 ∗ (5)  Mod(Eα ) + Mod(Fα ) + Mod(Gα ) ExtX (α) ∗  e f  g (6)  Log + + Log . τ ≥ τ0 for some uniform constant τ0 ). When the modulus of Fα is extremely small. which is determined in §7. the cylinder Fα in the middle and two expanding annuli Eα and Gα attached to the two boundaries of Fα . we use this idea to define a notion of extremal length for a saddle connection. The constant τ . Note that. qt ) = gt (X. Remark 4. the inner boundary of Eα (and similarly for Gα ). all statements hold with universal constants independent of τ as long as τ ≥ τ0 . Along Teichmüller geodesics. 1 . More precisely. f or g could be zero and the second line will be −∞.4. (re- call that 0 was chosen so that Lemma 4.1. l l l Where Log(  ) is a modified logarithm function:  Log(t) = max log(t). q). Later. let (Xt . the interior of some or all of these annuli could be empty. unless otherwise stated. This neighborhood can be decomposed into three annuli. Consider a q–geodesic representative of α and take the maximal regular neigh- borhood of α that is still an annulus. COUNTING CLOSED GEODESICS IN STRATA 12 c. The annuli Eα and Gα are called expanding because the equidistance curves parallel to the inner boundary get longer as they span Eα and Gα . Of course. Let l = `q (α) and let e. As in §2. Denote the set of short curves in q by Sq . the cases when α ∈ Sq is a cylinder curve and Fα has a large enough modulus will need special treatments. We use Mτ instead of just writing e−2τ to highlight the fact that Mτ is a bound for modulus.) We call the shared boundary of Eα and Fα . f and g be the q–distances between the boundaries of Eα . When ExtX (α) ≤ 0 .

Define the size sQ of Q to be the q–diameter of this representative. We call the components of S \ Sq the thick subsurfaces of q. Note that the q–geodesic representative of α is the saddle connection 1 (the end points of arc 1 are identified). The homotopy class of each such subsurface Q of S has a representative with q–geodesic boundaries. ExtX (α) ExtXt (α) ExtX (α) 4. The curve α is a non-cylinder curve and it has a small extremal length because the expanding annuli Eα and Gα (Fig. we ∗ have Modqt (Fα ) ≺ 1. Example 4. In fact. Assume that the edges 2.4. There are two thick subsurfaces. 3. For example. There is. The following theorem states that the geometry of the subsurface Q is essentially the same as that of the thick hyperbolic subsurface of X in the homotopy class of Q but scaled down to a size sQ : Theorem 4. the surface obtained from the polygon in Fig. As a consequence of Equation (5). The thick-thin decomposition of quadratic differentials. There is a once punctured torus with a bound- ary curve β whose q–representative is degenerate and is represented in q with a graph of area zero (the union of arcs 5 and 6).3 ([R2]). q) be the associated quadratic differential. For every essential simple closed curve γ in Q. we have 1 ∗ 1 ∗ 1 (7) −t≺ ≺ + t. A quadratic differential can be described as a singular flat structure of a surface plus a choice of a vertical direction. 2) has a large modulus. The other is a pair of pants whose . Assuming α ∈ Sq≤τ and 0 ≤ t ≤ τ . sQ In particular. 1 with the given edge identifications is a once punctured genus 2 surface. The curve β is a cylinder curve and has a small extremal length because the flat annulus Fβ (Fig. This can also be described as the smallest representative of Q with q–geodesic boundaries. Choose an arbitrary vertical direction and let (X. Sq = {α. 2) have large moduli (Sq≤τ = {α}). 3 4 2 5 β β 6 2 5 α α 6 1 1 4 3 Figure 1. a unique such representative that is disjoint from the interior of cylinders associated to the boundary curves of Q. β}. q) and short curves of X.4. Then the hyperbolic metric on X has two short simple closed curves. in fact. We denote this subsurface by Q as well. Quadratic differential (X. COUNTING CLOSED GEODESICS IN STRATA 13 where gt is the Teichmüller geodesic flow. ∗ ∗ `q (γ) p `X (γ)  ExtX (γ)  . β is a large-cylinder curve (Sq≥τ = {β}). the q–length of shortest essential curve in Q is on the order of sQ . 5 and 6 have a comparable lengths. the edge 1 is significantly shorter and the edge 4 is significantly longer than the others.

That is. If γ is contained in Q. The maximal expanding annuli Eα and Gα and the maximal flat annulus Fβ . Applying a surgery between γ and α0 (that is. for the resulting curve γ 0 we have ∗ `q (γ 0 ) ≺ `q (γ). we find a simple closed curve in Q of length comparable to (e + l). we note that. Proof. Using the notation of Equation (5) we have (1) l ≤ 2sQ . The previous argument ∗ now again shows (e + l)  sQ .3 we get. γ crosses Eα0 . (2) e ≤ sQ . ∗ (3) (e + l)  sQ . replacing a sub-arc of γ with the sub-arc of α0 with the same end points) will increase the length of γ by at most l0 and hence by at most a multiplicative factor. where γ exists by intersecting a curve α0 and returns by intersecting a different curve α00 is similar. if e ∗ ∗ ∗ Log  Mod(Eα )  1 then e  (e + l). Lemma 4. boundaries consist of two copies of α and one copy of β. The size of a thick subsurface Q is related to the radii of annuli Eα . we need to show (e + l)  sQ . l . Since α is part of Q. Hence. But α0 has small extremal length and.) Then. Let Q be a thick subsurface of (X. its length is less than twice the diameter of Q which is the first assertion. by Equation (5). The maximal expanding annuli Eα and Gα do not necessarily stay inside of the q–representative of this pair of pants and they may overlap. its outer boundary self-intersects. Fα and Gα for every boundary curve α. COUNTING CLOSED GEODESICS IN STRATA 14 Eα Gα Fβ Figure 2. This is a contradiction. Since Eα is maximal. Applying this argument to all boundary curves γ intersects. ∗ then ExtX (γ)  1 (Q is a thick subsurface). To prove part (3). To see part two. then Q is contained in Eα which is an annulus. Parts (1) and (2) ∗ ∗ imply (e + l) ≺ sQ . Fα0 and Gα0 and its length is larger than (e0 + f 0 + g 0 ). From Theorem 4. (The other case. We make a few observations that will be useful later. Then l + e  `q (γ). note that if e is larger than sQ . sQ If γ is not contained in Q we still can construct a simple closed curve in Q whose length is not much longer than γ.5. To see part (4). α be a boundary component of Q and Eα be the expanding annulus in the direction of Q. (e0 + f 0 + g 0 ) is much larger than l0 = `q (α0 ). q). ∗ ∗ (4) If Mod(Eα )  1 then e  sQ . `q (γ) ∗ ∗  1 and hence (e + l)  sQ . we can construct an essential simple closed curve γ as a concatenation of a sub-arc of α and two arcs ∗ connecting α to the boundary of Eα . Assume γ exits Q by intersecting a boundary curve α0 twice.

Extremal lengths and flat lengths of saddle connections. for qt = gt (q) we have 1 ∗ 1 ∗ 1 (8) −t≺ ≺ + t. If Extq (ω) is small enough. Denote the end points of ω by p1 to p2 .5. As mentioned above. . (4) follows from (3). Let l = `q (ω) and e be the radius of Eω (the q–distance between ω and the boundary of Eω ). Let qω be the lift of q to Xω . we need  to be small enough so that Lemma 4. that would provide roughly the same result. Using the notation of Equation (5). we have `q (γ) ≥ (e + f + g). There exists a unique ramified double cover φ : Xω → X with simple ramification points at only p1 and p2 . (e + f + g) ≥ l.6. By Equa- ∗ tion (5).8. making  smaller will not effect the constant involved in any of our estimates. since ExtX (α) is small. Hence. is to compute the extremal length in a double cover of (X. Proof. Note that αω = φ−1 ω is a simple closed curve on Xω . we can also define a notion of extremal length for saddle connections.  Mod(Eαω ) Mod(Gαω ) Mod(Eω ) Since l and e change at most exponentially fast along a Teichmüller geodesic. `q (γ)  (e + l). Hence. by Equation (5) ∗ 1 1 2 ∗ ExtXω (αω )  + =  Extq (ω). either • Extq (ω) ≤ .5. As we shall see. or • ω lies on a geodesic representative for α with ExtX (α) ≤ . However. Let α ∈ Sq be the boundary of a thick subsurface Q and let γ be any curve crossing α. let Ωq () be the set of saddle connections ω of q so that.13 and Lemma 7. Assume Extq (ω) is small enough so that Equation (5) holds for Xω . The final value of  will depend on τ . Then ∗ `q (γ)  sQ . knowing `q (ω) is small does not imply that ω has a small extremal length. Note that αω has a unique geodesic representative in qω (Mod(Fαω ) = 0) and Eαω and Gαω are conformally equivalent to Eω . then ∗ ExtXω (αω )  Extq (ω). COUNTING CLOSED GEODESICS IN STRATA 15 Now.1 follows from Theorem 1. The corollary now follow from part (3) of Lemma 4. we have the following lemma which is enough to show that The- orem 1.  As a corollary we get the following analogue of the collar lemma: Corollary 4. Let Eω be the annulus obtained by taking the largest regular neighborhood of ω that is still a topological disk and then cutting a slit open along ω. Let ω be a saddle connection connecting two distinct critical points in (X. Lemma 4. Then. However. Proof. similar to Equation (7).1 hold. Log(e/l) Mod(Eω ) Another interpretation of this notion of extremal length. q). q).5.7. For any  > 0. we define 1 ∗ 1 Extq (ω) =  . In general. Extq (ω) Extqt (ω) Extq (ω) Definition 4.  4.

there is always a thick subsurface of size comparable to 1. . we have (recall that s0 = l1 ≺ (e + l)): k !   Y si e+l ∗ 1 e+l ∗ 1 (9)   . Qk be a sequence of distinct subsurfaces of sizes s1 . Let Q1 . . the maximum value of k depends only on the topology of S. Note that in both Lemma 4. q) has a saddle connection ω with `q (ω)  1. 1  Log  Log .5. we know that   1 ∗ gi + gi + li + si  max Log . we conclude that either 1 ∗ 1 e+l ∗ 1 there is some i where.5 implies. Ωq () is non-empty. Namely. sk respec- tively. we get: ∗ ∗ ∗ (e + l)  `q (γ)  sQ  `q (α1 ). either γ ∈ S and we can take α1 = γ or γ intersects a thick subsurface Q in which case we let α1 be any boundary component of Q. taking the logarithm of both sides of Equation (9). Extq (αi ) l l l In the first case. This is because. since ω is assumed to be short.9 the implied constants only depend on the topology of S. where α1 is a boundary component of Q1 . s i=1 i−1 l s0 l l Here. obtained by a concate- nation of a sub arc of ω and two arcs connecting ω to the boundary of Eω . with ∗ ∗ `q (γ) ≺ (e + l). Let l = `q (ω) and e be the radius of Eω . Since the boundary of Eω self intersects (Eω is maximal). . if α1 does not have small q–length then Extq (ω) is small and if α1 does have a small q–length then two adjacent surfaces ∗ have to have very different sizes. . We will argue that.6 and part one of Lemma 4.  Extq (ω) l l `q (ω) Remark 4. the expanding annulus with inner boundary αi in the direction of ∗ Qi with radius gi . Therefore. Consider Gαi . Hence.10. Extq (αi ) li li si−1 That is. .  Log or Log  Log . . (gi + li )  si and by ∗ part one si−1  li . Since the total area of q is 1. from Equation (5). . Proof. In particular. Using Corollary 4. the common boundary curve of two surfaces of very different size has a very small extremal length.9. 1 ∗ e ∗ e+l ∗ 1  Log  Log  Log . there is a simple closed curve γ. Assume that (X. Let li = `q (αi ) and let s0 = l1 . For i ≥ 1. In second case. we have e + l is much larger than l. the lemma holds for α = αi .7 and Lemma 4. either 1 ∗ 1 1 ∗ 1  Log or  Log . This implies that there is a curve α1 ∈ S with `q (α1 ) ≺ (e + l). Qi−1 and Qi share a boundary ∗ curve αi and sk  1. That is. Then. part (2) of Lemma 4. COUNTING CLOSED GEODESICS IN STRATA 16 Lemma 4. Extq (ω) `q (ω) ExtX (α) `q (ω) for some simple closed curve α. . .

the complement of T is a union of triangles and large-cylinders Fα◦ . These curves divide the surface into subsurfaces with nearly geodesic boundaries.11 large enough so that every quadratic differential q has a (q.13.11. For every τ there is 1 (τ ) so that for  < 1 (τ ) the following holds. However. let υα be an arc connecting the boundaries of Fα that is perpendicular to α. denote the interior of a cylinder Fα by Fα◦ . In particular. τ )–regular triangulation T of q we mean a collection of disjoint saddle connections satisfying the following conditions: (1) For α ∈ Sq≥τ . Given a cylinder curve α. q) be a quadratic differential. as mentioned before. we perturb α slightly to a curve α that is a union of saddle connections. A (q. We denote the surface associated to Q with Q. Remark 4. COUNTING CLOSED GEODESICS IN STRATA 17 4. By a (q. saddle connection may intersect more than one thick subsurface. α ∈ Sq≥τ . α ∈ Sq≥τ That is. Then Ω can be extended to a (q. We then extend Ω to a triangulation in each Q so that the edge lengths are not much longer than the diameter of Q which comparable to sQ (see Claim 3) and let T be the union of these triangulations. we would like the q–length of edges to not be longer than the thick subsurfaces they reside in. We would like to mark a quadratic differ- ential q by a triangulation where the edges have bounded length. Then. otherwise the resulting triangulation would not be (q. τ )–regular triangulation. In fact. Recall that two saddle connections are said to be disjoint if they have disjoint interiors but they may share one or two end points. saddle connections in Ω may intersect a boundary curves α of Q. τ )– regular triangulation. However. (3) If α is a cylinder curve in Sq≤τ then υα intersects T a uniformly bounded number of times. That is. We would like to triangulate each thick piece Q separately and let T be the union of these triangulations. Hence. However. To remedy this. τ )–regular. . Let (X.12. It is important to choose the implied constants in conditions 2 and 3 in Definition 4. Of course. T contains the boundaries of Fα . We shall see that condition 3 means that the triangulation T does not twist around short simple closed curves. one needs to be careful that Q does not intersect any subsurface of size much smaller that sQ .6. we triangulate only the compliment of large-cylinders. ∗ (2) If an edge ω of T intersects a thick subsurface Q of q then `q (ω) ≺ sQ . the notion of having a bounded length should depend on which thick subsurface we are in. Let Ω be a subset of Ωq () consisting of pairwise disjoint saddle connections. lies in a small neighborhood of α and is disjoint from Ω (see Claim 1). Definition 4. Lemma 4. Also. T is disjoint from Fα◦ and it triangulates their complement [ q\ Fα◦ . large-cylinder curves will require a special treatment.13 below holds. τ )–regular triangulation T . we choose the constants so that the key Lemma 4. Proof.

α and β do not intersect.5 and the previous assumption we have ∗ ∗ e  sQ  sR . we can take α = α. COUNTING CLOSED GEODESICS IN STRATA 18 Claim 1: For every α ∈ Sq . This is because if ω is an arc in a curve β ∈ S it does not intersect α since β and α have intersection number zero. β ∈ Sq .5). Then α is in the homotopy class of α and lies inside E α . we can ensure that α is disjoint from Ω. Let e be the distance between the boundaries of Eα . Assume Eα . Denote the (`q (α)/2)–neighborhood of α in Eα with E α . 1 Extq (ω) `q (ω) Mτ which is a contradiction. by the triangle inequality. But E α does not intersect any small subsurfaces. the last condition of the claim is satisfied as long as α stays in E α . Consider the the union of α and all the arcs in Ω that intersect α. Extq (ω) is small. if ω is part of a short curve α0 . For α. has a large modulus. namely. But the injectivity radius of any point in E α is less that 2`q (α). and hence: ∗ ∗ sQ0  `q (∂Q0 ) > e − `q (α)/2  sQ . This is because. since α is already disjoint from Ω. Also. Furthermore. again ω is disjoint from α because short curves α and α0 do not intersect. . The convex hull of this set in E α is an annulus (perhaps degenerate) and has two boundary components. then either Eα or Gα has a large modulus. any saddle connection ω that appears in α has a q–length less than or equal to 2`q (α). (by choosing  small enough) `q (ω) is much less than (with a multiplicative factor as large as we like) the distance between the boundaries of E α with is equal to `q (α). But Mod(Fα ) = `qs(α) α ≥ Mτ and thus 1 1 eω 1 ≤ = Log ≤ Log = 2τ. if α is a boundary of Q then α intersects only surfaces that are larger than Q. The annulus Eα may not be contained in entirely Q and may intersect some thick subsurfaces with very small size. if α intersects a thick subsurface Q0 we have: ∗ sQ0  sQ . This proves that even a concatenation of arcs in Ω cannot cross E α because the total number of arcs in Ω is uniformly bounded. by choosing 1 < 1/2τ . Fα does not contain any singular points and any arc ω ∈ Ω intersecting α has to cross Fα . lies in a (`q (α)/2)–neighborhood of α and is disjoint from Ω. Since Q0 is disjoint from α. Thus. To see this. If Mod(Fα ) ≥ Mτ . In this case (Mod(Fα ) ≥ Mτ ). Let α be the boundary component that is not α. the annulus in the direction of Q. we have eω ≤ `q (α). `q (ω) ≥ sα and. assume Q0 intersects E α . By part (4) Lemma 4. But e is much larger than `q (α). for the radius eω of Eω . Proof of Claim 1: Let α ∈ Sq be the common boundary of a thick subsurfaces Q and R. Otherwise. Note that. there is a representative α of α that is a union of saddle connections. it has to enter Eα intersecting the outer boundary of Eα . Hence. which implies that its length is much less than the injectivity radius of any point along ω. Then. If Mod(Fα ) ≤ Mτ . Note that no arc in Ω can cross E α (intersect both boundaries). The annulus with the larger modulus is in the direction of the thick surface with the larger size (Lemma 4.

we obtains a collection of subsurfaces with nearly geodesic boundaries. Assume `q (α) ≥ `q (β). for ω to intersect E α it has to intersect α. β ∈ S. Since α is disjoint from β. Claim 3: A partial triangulation of Q where the length of edges are less than a fixed multiple of sQ can be extended to a triangulation using saddle connections of length less than a larger fixed multiple of sQ . We can pick this saddle connection to have a length comparable to that of α (the distance between α and α is comparable to the length of α and the both contain singular points) which is smaller than sQ or the size of any subsurface Q intersects.  The next claim describe how to add the remaining edges to T0 to satisfy condi- tions (1) and (2). The radius of Eω is much larger than length of ω. For each α ∈ Sq≤τ where Fα intersects Q we add a saddle connection ωα that crosses Fα and does not twist around it. which implies Eω contains Q. then `q (ω) ≥ `q (α). These edges satisfy conditions (1) and (2) and adding them. we need to show that ω intersects υα a bounded number of times. This finishes the proof of claim 1. Hence. if they intersect more than once. We will show that α is disjoint from any saddle connection ω ∈ Ω that intersects β. This is because ω is shorter than β and hence shorter than α. After removing the interiors of large cylinders from the quadratic differential (X. Hence. Also. then β in contained completely in Eα which is a contradiction. α and β are disjoint. Denote the representative of a thick subsurface Q by Q. we are guaranteed that the final triangulation T0 satisfies the condition (3). in the end. Proof of claim 3: We prove the claim by induction. Namely. We show that arcs in T0 satisfy condition (3). this would show that α is also disjoint from β. To see that an arc ω ∈ Ω satisfies condition (2) note that if it did not. if ω ∈ Ω intersects a cylinder Fα . α ∈ Sq . each of which intersect υα at most once. α ∈ Sq≤τ . But then Eω would contain the curve α which is a contradiction (Eω is a topological disk). Claim 2: T0 = Ω ∪ Ω ∪ Ωn is a partial triangulation of Q satisfying conditions (2-3) of Definition 4. the curve α is a convex hull of the union of the curve α which is disjoint from Fα and a bounded number of arcs in Ω. We have already shown that these edges satisfy condition (1) and arcs in α satisfy condition (2). but the argument is long since we have to look at all the cases. Assume ω intersects β and E α . Then it is disjoint from α by the construction of α. we have several components each with diameter .  Next. q) and cutting along curves α. Each cutting increases the diameter by at most twice the length of edge being cut. Let Ωn denote the set of saddle connections ωα . If the the outer boundary of E α is close to β. In fact. ω would intersect a thick subsurface Q with `q (ω) ≥ sQ . This is a contradiction. COUNTING CLOSED GEODESICS IN STRATA 19 It remains to show that for α. Hence α intersects υα at most a bounded number of times and thus arcs in α satisfy condition (3). Start by cutting Q along the given edges. The arc ω can not intersect the outer boundary of E α . Proof of Claim 2: All the conditions follow immediately from the construction.11. let Ω be the set of edges that appear in curves α for every α ∈ S.

but any two such transverse arcs have bounded geometric intersection number (see [Mi]) and the associated relative twisting twistα (  . γ̃). τ )–regular. The diameter grows multiplicatively each time but still it is uniformly bounded multiple of sQ in the end.  ) is determined up to a uniformly bounded additive error. In the simplest case.7. Let S̃α be the annular cover of S associated to α and denote the core curve of S̃α again by α. The claim follows from the fact that the this process ends after a uniformly bounded number of times. Now consider a more general case where α is a curve on the surface S and β and γ are two transverse curves to α.  4. homotopy is relative to the end points of an arc). The relative twisting between two structures is the relative twisting between the associated arcs in S̃α .  ). the choice of β̃ is not unique. We choose the constant in the second condition of a (q. γ) = twistα (β̃. Let β̃ and γ̃ be the lifts of β and γ to S̃α (respectively) that connect the boundaries of S̃α . This is essentially the definition introduced by Minsky extended to a slightly more general setting. This is often only coarsely defined. that is. Alternatively. condition (2) of being (q. Then we say the given structure defines a notion of zero twisting around α. In this section we define several notions of twisting and discuss how they relate to each other. Then β̃ can be taken to be the geodesic in X̃α that is perpendicular to α in the Poincare metric of X̃α . τ )–regular is still satisfied. . the value of twistα (  . We can generalize this further and define twisting between any two structures on S as long as the structures in question provide a (nearly) canonical choice of a homotopy class of an arc β̃ connecting the boundaries of S̃α . we are done. however any pair of lifts are disjoint. We denote the relative twisting of two objects or structures around a curve α by twistα (  . Hence. let A be an annulus with core curve α and let β and γ be homotopy classes of arcs connecting the boundaries of A (here. some component contains a saddle connection that is not part of its boundaries or the given triangulation. The relative twisting of β and γ around α. The arc β̃ is not uniquely defined. Twisting and extremal lengths. which is comparable to sQ . Otherwise. X) is well defined up to an additive error. γ). we can pick the shortest curve β transverse to α and let β̃ be the lift of β that connects the boundaries of S̃α . Since the newly added edged in Q have a length less than sQ and any thick subsurface Q intersects has a size less than sQ . Here are a few examples: • Let X be a Riemann surface.  The triangulation T is now the union of all the saddle connections in T0 and those coming from claim 3. is defined to be the geometric intersection number between β and γ. This is well defined up to an additive error of 2 (see [Mi]). In any case. Note that freely homotopic curves lift to arcs that are homotopic relative their endpoints. COUNTING CLOSED GEODESICS IN STRATA 20 comparable to sQ . the resulting triangulation T is (q. We now define twistα (β. The shortest such saddle connection has a length less the diameter of the component it is in. τ )–regular triangulation large enough so that the outcome of this algorithm is in fact a (q. twistα (β. τ )–regular triangulation. using the previous case. If all components are triangles.

X) ExtX (α) + max i(γ. COUNTING CLOSED GEODESICS IN STRATA 21 • Let q be a quadratic differential. α)2 twist2α (γ. Since the number of elements in SX and BX is uniformly bounded. τ )–regular triangulation of (X. then we choose β to be a curve that intersects γ once or twice. q). As before. then the corollary holds for β = α (the second conclusion follows from the fact that a short curve does not twist much around a long curve). The expression “fix a notion of zero twisting around α” for a curve α in S means “choose a homotopy class of arcs connecting the boundaries of S̃α . i(γ. Then we can choose a curve β transverse to α that is carried by T and has a bounded combinatorial length in T and let the lift of β to the annular cover of α define zero twisting. In this section we establish some statements relating Extremal length. If γ is not short in X. the associated relative twisting twistα (  . and let SX be a set of 0 –short simple closed curves in X and let BX be the set of simple closed curves of bounded length in X. X) ExtX (α) or i(γ. ∗ i(γ. there is a curve β so that.14 applies to γ. α) = O(1) for α ∈ BX .   ∗ 1 (10) ExtX (γ)  max i(γ. is disjoint from other curves in SX .14 (Minsky). β̃ can be taken to be the geodesic in q̃α that is perpendicular to α in the Euclidean metric coming from q or a lift of the q–shortest curve β transverse to α (see [CSR]). • Let T be a (q. Note that the reverse of first inequality always holds (Equation (2)). Since curves with bounded combinatorial length intersect a bounded number of times. is again well defined up to an additive error. q) and α ∈ Sq≤τ . X) is large (That is. ExtX (α) In the fist two cases α ∈ SX and in the third case α ∈ BX . ∗ We say β is a curve of bounded length in X if ExtX (β)  1.α)2 If ExtX (γ) ≺ Ext X (α) .8. β) and twistγ (X. α)2 . Theorem 4. for α ∈ SX . ∗ p ExtX (γ) ExtX (β) ≺ i(γ. Suppose X ∈ T (S). Theorem 4. twisting and intersection number. α)2 + twist2α (γ. if β twists around α a lot). We start with a theorem of Minsky giving an estimate for the extremal length of a curve. Then given a simple closed curve γ 6∈ SX . where twistγ (β. It follows from the definition of twisting and elementary hyperbolic geometry ∗ that if twistα (β. X) is bounded and where ∗ 1 i(β. α)2 . . α∈SX ExtX (α) α∈BX The multiplicative constant depends only on the topology of S. Proof.15. ExtX (γ) is comparable to one the following terms: i(γ. If γ ∈ SX . α)2 . We argue in 3 cases. Applying Equation (10) to β we have ExtZ (β)  ExtX (γ) which implies that the corollary holds for β and γ. T ). Corollary 4.” 4. Intersection and twisting estimates. We denote the associated relative twisting with twistα (  . β) = O(1). For every curve γ and any X ∈ T (S). then ExtX (β)  ExtX (α).

∗ ∗ The curve γ also intersects α and hence ExtX (γ)  ExtX1 (α)  ExtX (β). . α)2 twist2α (γ. ∗ (Assumption on γ) Ext(γ) ≺ i(γ.16 (Rafi). β) times (up to an additive error). s Proof. τ )–regular triangulation is supposed to capture the geometry of q. Let s be minimum of sQ where Q is a thick subsurface of q that intersects ωT . Theorem 4. Hence. In the first lemma. Thus. Therefore. every strand of γ intersecting α intersects β at least twistα (γ. In particular + (11) twistα (γ. Condition (2) in the definition of a (q. β does not twist much around γ. X) = O(1). X) ExtX (α) ∗ i(γ. ω) ≺ + 1. we assume `q (ωT ) ≤ s/3. α)2 . without loss of generality. twistα (ω. Y ) = O(1). ∗ twistγ (β. β) (Equation (11)) ≺ ExtX (β) ∗ i(γ.9. X)  twistα (γ. β) ≺ i(γ. β)2 (Equation (12)) ≺ . β). q) and Riemann surface Y ∈ T (S) with dT (X. The second lemma shows that the notion of zero twisting coming from q or T is the same. the corollary follows. we take β = α. Again. we relate the length of a saddle connection to its intersection number with a (q. COUNTING CLOSED GEODESICS IN STRATA 22 In the second case. Then ∗ `q (ω) i(ωT . we have ∗ 1 twistα (Y. ∗ The last case is when α ∈ SX and ExtX (γ) ≺ i(γ.17. Let ω be any other saddle connection in q so that. for every curve α ∈ Sq≤τ . As we mentioned at the beginning of the section. τ )–regular triangulation. β) is large and the additive error can be replaced by a multiplicative error to obtain ∗ (12) i(γ.  We also recall the following lemma ([R1. In this case. q) = O(1). τ )–regular triangulation and ωT be an edge of T . X) = O(1) and ExtX (β)  1. For quadratic differential (X. α) twistα (γ. In this case twistα (γ. q) ≺ . τ )–regular triangulation implies that ∗ `q (ωT ) ≺ s. we take β to be a curve transverse to α with (see above) ∗ ExtX (β)  ExtX1 (α) and twistα (β. It is sufficient to prove the lemma for a subsegment of ωT with a q–length less than s/3. ExtX (α) 4. Since β has bounded length in X. β). ExtX (β) which implies the corollary. a (q. Geometry of quadratic differentials and (q. τ )–regular triangulations. Let T be a (q. because ωT can be covered but uniformly bounded number of such segments. We make this explicit in the following two lemmas. Also.3]): Lemma 4. Lemma 4. α)2 twist2α (γ.

ω Q3 Fα1 N ωT Fα2 Q1 Q2 Figure 3. it can not be an essential curve in any subsurface Q that ωT intersects. the total intersection number outside of cylinders Fα is uniformly bounded as well. Q2 and Q3 . Hence. α2 ∈ Sq≤τ and thick subsurfaces Q1 . (See [CSR. we need to show that i(ωT . COUNTING CLOSED GEODESICS IN STRATA 23 Consider the s/3–neighborhood N of ωT . ω̄) = O(1). τ )–regular triangulation and the twisting assumption on ω. it may intersect at most two such curves. Lemma 5.  . (This follows from the definition (q. However. Otherwise.) As before.) Since the number of connected components of N \ (∪Fα ) is uniformly bounded. any non-trivial curve in N is homotopic to some curve in Sq . and ωT and ω̄ can intersect at most once in each component. The arc ωT intersects curves α1 . This finishes the proof. It remains to show that the number of intersection points outside of all cylinders Fα is abounded. at least one of them has to twist around Fα a large number of times. the saddle connection ωT may intersects some curve α ∈ Sq≤τ . Each component of ω restricted to N intersect ωT only a bounded time outside of cylinders Fα1 and Fα2 . Note that. The number of intersection points inside of Fαi are bounded because of the assumption on the twisting. Every component ω̄ of ω ∩ N that intersects ωT has a q–length of at least 2s/3. it has to be homotopic to some curve α ∈ Sq .) If two arcs inside of a cylinder have a large intersection number. Hence. (In fact. To see this we observe that it is not possible to have a subsegment of ωT and a subsegment of ω̄ that are disjoint from cylinders Fα and have the same endpoint. by the defini- tion of s. This is because both ωT and ω̄ intersect να a uni- formly bounded number of times.6] for a more detailed discussion. Which means. the concatenation would create a two segment curve β that is non-trivial in N . First we observe that the number of intersection points between ωT and ω̄ inside of Fα is uniformly bounded. α and β create a cylinder with total negative curvature which contradicts the Gauss-Bonnet theorem. let να be an arc in Fα that connects the boundaries of Fα and is perpendicular to them. This is because. any nontrivial loop in N has a q–length of less than s and.

τ )–regular triangulation T we have twistα (T. q). j 6= i. For a quadratic differential (X. Consider such a representative traversing the minimum possible number of edges. Define [ M= µQ ∪ S0 ∪ {βα }α∈S0n . where the curves in M and T have representatives with the following properties: (1) curves in M have no self intersections and intersect each other minimally. τ )–regular triangu- lations T on a quadratic differential (X. τ )–regular triangulations on (X. T is equivalent to φ(T 0 ) where φ is a multi-twist with support on curves in S0 . Also. q) is equivalent to a q 0 –regular triangulation −1 −1 T 0 on (X 0 . That is. by definition. Now. q) where X ∈ U is uniformly bounded. We start with a topological counting statement. For a multi-curve S0 . We now count the number of (q. q) = O(1). The number of (q. On the other hand. τ )–regular Triangulations near a point in Teichmüller space. Then the number of equivalence classes. and let β be an essential curve in Q1 ∪ Fα ∪ Q2 that is transverse to α and has a bounded combinatorial T –length. γk } for Q together with a transverse curve γ i for 0 ≤ i ≤ k. Proof. That is. . intersects γi once or twice and is disjoint from γj . up to twisting around SX0 . q 0 ) if the pre images fX (T ) and fX 0 0 (T ) are homotopic on S. for α ∈ S0n . Its restriction to Qi has a length bounded by O(sQi ) and its restriction to Fα has a length bounded by `q (α) (Mod(Fα ) is bounded and there is no twisting around α) which is less than both sQ1 and sQ2 . Let U be a ball of radius one in T (S) centered at X0 . α ∈ Sq≤τ and a (q. Q Claim: Give a set M as above. τ )–regular triangulation T on (X. A representative for the curve β can be constructed using edges of T that intersect either Q1 or Q2 . i(β. q) as topological objects on S. triangulating S \ S0c . . Let S0 = S0c ∪ S0n be a system of curves on S For every subsurface Q in S \ S0 . Proof.18. µQ is a pants decomposition {γ1 . Let γ be a curve transverse to α with the shortest q–length. we say a (q. let βα be a curve transverse to α that is disjoint from all other curves in S0 and i(βα .17 implies that γ intersects any edge of T at most a bounded number of times. we say T is equivalent to T 0 up twisting around S0 if. . of (q. Then γ intersects α once if Q1 = Q2 and twice otherwise. µQ ) = O(1). after being pulled back by the marking map fX : S → X. γ) = O(1). q) glued along the cylinder Fα . each edge of T appears at most twice along the representative of β. the total number of edges of T is bounded by the topology of S. Lemma 4. Hence. otherwise a surgery argument would reduce the length of β.  4. . The homotopy does not have to fixed the vertices of T . We can think of a (q. τ )–regular trinagulations. Each γ i is contained in Q. there is a uniformly bounded number of possibilities for the homotopy class of a triangulation T . up to homotopy.10. q)  i(β. Let Q1 and Q2 be the thick subsurfaces of (X. COUNTING CLOSED GEODESICS IN STRATA 24 Lemma 4. Also.19. γ) is uniformly bounded. An argument similar to that of Lemma 4. let µQ be a marking for the subsurface Q in the sense of [MM]. + twistα (T. This finishes the proof. .

Hence. q). τ )–regular triangulation Tq on (X. Curves in µQ have a uniformly bounded length on X and hence a uniformly bounded length in X0 . that for every (q. the total number 0 of choices for transverse curves is of order of G(X0 ). The first part of condition (4) is a consequence of Lemma 4. (4) for α ∈ S0n .18 and the second part again follows from Lemma 4. α) = 0. intersect each other minimally and the intersection number with T does not increase. twistα (T. Notation.17. It remains to show. First we need to establish some notations. i(T. up to twisting around SX0 there are only finitely many choices. τ )–regular triangulation Tq in (X. Now for each α.  5. we establish some intersection bounds for regular triangulations associated to quadratic differentials that appear at the end points of a geodesic segment in Q(σ). there is a uniformly bounded number of possibilities for the location of vertices of T (up to homotopy). Let (X. We construct Mq as follows: The curves Sq = Sq≥τ ∪Sq≤τ have a bounded length in X0 hence there is a uniformly bounded number of possibilities for these sets. Condition (2) follows from the construction of (q. Once the vertices of T are fixed. τ )–regular triangulations. there is a set of simple closed curves Mq so that Tq and Mq satisfy the above properties and then to bound the number of possibilities for the set Mq . that can appear as an edge of T . let βαq be the shortest q transverse ∗ curve to α. For each thick subsurface Q of q. Perturb the q–geodesic representative of curves in Mq so that they have no self-intersections. each region is either a polygon where there is a unique arc (up to homotopy) connecting any two edges (or a vertex to an edge) or an annulus neighborhood of a curve α ∈ S0c where there are two possibilities (edges of T are simple and disjoint from curves in S0c . γ) = O(1). However. Q Then the total number of possible sets Mq chosen as above is of the order of G(X0 ).1. q) where X ∈ U . This is because there is a bounded number of possibilities for the intersection pattern of the arc with the complementary regions. with end points on these vertices. note that the curves in M divide S into bounded number of complementary regions. we need to check that the conditions (1)-(4) hold for Tq and Mq . each one is either a polygon or an annulus parallel to a curve α ∈ S0c . Lemma 4. Also. . βα ) = O(1). Condition (3) follows from Lemma 4. 5. βαq ) ≺ ExtX1 (α) . Hence there are only a finely many choices for these as well. q) be a quadratic differential so that X ∈ U .17. To see the claim. For a (q. α) = O(1). (3) for any γ ∈ µQ .16 implies that twistα (X0 . In this section. Define [ Mq = µQ ∪ Sq ∪ {βαq }α∈Sq≤τ . choose a q–short marking µQ in Q. let Q(σ) be the stratum of quadratic differentials with of type σ. Intersection bounds between regular triangulations As before. i(T. Hence the number 0 of possible choices for βαq is of the order of ExtX1 (α) and therefore. there is a uniformly bounded number of possibilities for an arc. and i(T. COUNTING CLOSED GEODESICS IN STRATA 25 (2) for any α ∈ S0c .

2. then twistα (qa . Hence we can consider Tb as a union of saddle connections in qa . τ ) to be the set of points Z ∈ T (S) so that there is a Teich- müller geodesic GZ : [a. We say ωb intersects ωa positively. τ )–regular triangulation and Tb be a (qb . b] → Q1 T (σ). such that dT (Xa . τ ) as a ball of radius τ centered at X. ωb ) = O(1) and twistα (qa .1. the slope of ωb is larger than the slope of ωa . ωa ∈ Ta and αa ∈ SX . except that one is allowed only to move in the direction of Q(σ). As a consequence of Equation (5) we have a a 1 ∗ d 1 ∗ f (13) ≺ a and  a. Definition 5. `qb (α) . let la = `qa (α) and let da = ea + f a + g a be the distance between the boundaries of the union Aaα = Eαa ∪ Fαa ∪ Gaα . by definition of Sq≤τ . qt ). COUNTING CLOSED GEODESICS IN STRATA 26 1. Then Ta and Tb have identical vertex sets and their edges are either identical or intersect transversally. We use the notation of Equation (5) for qa and denote the flat and expanding annuli associated to a curve α by Eαa . dT (Xb . b − a ≤ τ. Define B(Q(σ). The slope of a saddle connection in qa (or in qb ) is a well defined number in the interval [0. qt ) ∈ Q(σ). Let ωa be a saddle connection in qa and let ωb be a saddle con- nection in qb . ωb ∈ Tb and αb ∈ SZ . let α ∈ Sqb . Fαa and Gaα and distances between their boundaries by ea . Z) = O(1). ExtX (α) l ExtX (α) l 3. ∞]. `qa (α) Therefore `qb (να ) ∗ ≺ 1. f a and g a . Proof. if when considering them both in qa (or qb ). Let να be the arc connecting the boundaries of Fαa and is perpendicular to them. One could think of B(Q(σ). Then. Let Ta be a (qa . 5. ωa ) = O(1) and twistα (qb . 2. τ )–regular triangula- tion. and (Xt . then twistα (qb . X. Intersection and twisting bounds between Ta and Tb . X. The geodesic flow induces a one-to-one correspondence between saddle con- nections of qa and qb .2. ≤τ Similarly. αb ) = O(1). Let υa be an arc of length f a connecting the bound- aries of Fα . We use similar terminology for intersection between a saddle connection and a cylinder curve and two cylinder curves. we assume that qa and qb . For the rest of this subsection. αa ) = O(1). GZ (t) = (Xt . Ta and Tb are as described in the beginning of the section. Let α ∈ Sq≤aτ . Also. We say ωb intersects ωa essentially positively if either ωb intersects ωa positively or i(ωa . ωb ) = O(1). X) = O(1). `qa (να ) ∗ = Mod(Fαa ) ≺ e−2τ . Lemma 5.

we say T intersects α essen- tially positively if any saddle connection in T intersects any saddle connection in the geodesic representative of α essentially positively. . The proofs of the other two assertions are similar.3. ωb )2 ≺ sa sb 2τ ∗ xa yb e ∗ xb ya ≺ ≺ = O(1). Also. Then ωb intersects ωa essentially positively and ∗ i(ωa . Lemma 4. Let Qb . COUNTING CLOSED GEODESICS IN STRATA 27 That is. A short curve can not twist around any other curve. we can multiply the two inequalities ∗ and take a square root to get i(ωa .17 implies that ∗ `q (ωa ) i(ωa . Otherwise. in view of Remark 5. we have ∗ `qa (ωa ) ≺ sa .  sa sb sa sb For a simple closed curve α and a triangulation T . ωb ) ≺ eτ . sb sb Similarly. Now assume that ωb does not intersect ωb positively. But the same is true for ωb . since i(ωa . The length of ωa in Qb is ∗ p (xa eτ )2 + (ya e−τ )2  xa eτ + ya e−τ . ωb ) is large. τ )–regular triangulation. by the definition of a (qa . xb and yb be similarly defined. ωb ) ≺ eτ . Let Qa be the thick subsurface of qa with the smallest size that intersects ωa and let sa be the size of the subsurface Qa . We denote the horizontal and the vertical lengths of ωa by xa and ya . να twists around α in qb a bounded number of times.4. This means that the slope of ωa in qb is larger than the slope of ωb . sb . ωb ) ≺  sa sa ∗ ∗ Observing that xa .17 is satisfied and can be applied freely. xa eτ xb From the product of inequalities in Equation (14) and Equation (15). ωb ) = O(1) we are done. Hence i(αb .  Remark 5. sb ` (ω ) However. αb ) is uniformly bounded. That is ya e−τ yb ≥ =⇒ xa yb e2τ ≤ xb ya . Proof.3. This gives a bound on i(ωb . Let ωa and ωb be edges of Ta and Tb respectively. Recall that. yb ≺ sb . ωb ) ≺ b + O(1). ya ≺ sa and xb . considering ωa and ωb in Qa we get ∗ `qa (ωb ) ∗ xb e−τ + yb eτ (15) i(ωa . ωb ). Which means twistα (qa . the curve αb is short in Z and hence in qb . Lemma 5. να ) is uniformly bounded. `qb (ωa ) ∗ xa eτ + ya e−τ ∗ (14) i(ωa . That is. ωb ) ≺  . να ) and thus on twistα (qa . we have ∗ xa xb + ya yb + xa yb e2τ + xb ya e−2τ i(ωa . If i(ωa . The main consequence of this lemma is that the twisting condition of Lemma 4. qbsb a needs to be large and we can incorporate the additive error into the multiplicative error. Considering ωa and ωb in Qb .

we obtain ∗ p i(α. τ )–regular trinagulation). we get `qb (α) ∗ la (17) i(α. Denote this cylinder by Fαb . Tb ) and twistα (qa . We need to consider the image of Fαa in qb under the Teichmüller geo- desic flow. Tb ). Tb ) by finding separate upper bounds for twistα (qa . qb ) i(α. . Applying Lemma 4. ωb ) ≤ ≺ e a. ExtX (α) Let ωb be an edge of Tb and let Qb be the thick subsurface of qb with the smallest ∗ size intersecting ωb . ExtX (α) (The term ExtZ1 (α) is omitted from the right hand side because it is bounded and can be absorbed in the multiplicative error. Let lb = `qb (α). We now find an upper-bound for twistα (qa . Tb ). ExtZ (α) Proof.17 to ωb and α in qb . Dividing both sides by ExtX (α) we obtain i(α. taking the square root we and summing over all arcs in Tb we get r ∗ τ la i(α. to prove the lemma. Let α be a simple closed curve in SX \ SZ . qb ) i(α. Let sb be the size of Qb (thus `qb (ωb ) ≺ sb by the definition of a (qb . COUNTING CLOSED GEODESICS IN STRATA 28 Lemma 5. da In view of Equation (13). Z) ≺ twistα (qa . Tb ) ≺ p . Tb ) ∗ eτ + O(1) ∗ eτ (18) ≺p ≺p . Tb ) − O(1) ≺ eτ ExtX (α). qb ) + . ExtX (α) ExtX (α) ExtX (α) This is the first estimate we required. Z) i(α. Z) i(α. it is τ sufficient to show that the expression √ e is an upper-bounds for both ExtX (α) i(α. if α ∈ SZ and α 6∈ SX then α intersects Ta essentially positively and ∗ eτ twistα (X. Tb ) is somewhat similar to above. Hence. a `qa (ωb ) ∗ τ sb i(α. each component of ωb ∩ Aα has a length larger than da . If α ∈ SX and α 6∈ SZ then α intersects Tb essentially positively and ∗ eτ twistα (X. sb sb Also.16 to the pair X and qa and to the pair Z and qb . but the two bounds do not imply each other. The argument involved in this new upper bound for i(α.5. da d Multiplying these two equations. that is la f a = lb f b . Note that the area of Fαa and Fαb are equal. Applying Lemma 4. we get ∗ 1 (16) twistα (X. qb ) and i(α. Tb ) − O(1) ≺ e . Ta ) ≺ p . ExtX (α) Similarly. ωb ) − O(1) ≺ ≤ eτ .) Hence. the distance between its boundaries by f b and let υαb be an arc of length f b connecting the boundaries of Fαb .

taking the square root and summing over all arcs in Tb . ExtX (α) The estimate in the Lemma follows from Equations (16). Then Corollary 4. υαb ) ≤ a . Thus ωb intersects α essentially positively. qb ) i(α. the relative twisting of qa and qb around α is comparable to the intersection number between υαa and υαb which is at most the qa –length of υαb divided by the qa –length of α. That is ∗ eτ f b twistα (qa . ωb ) ≺ = a a. Hence. Then the qb –length of every component of ωb ∩ Fαb is larger than f b . If the slope of ωα was smaller ∗ than ωb (say in qb ) then we would also have `qa (ωa )  `qa (ωb ). fb l f As before.6. Let ωa be a saddle con- nection of α that intersects ωb many times. l Taking a product and using the second part of Equation (13) we get:  b b ∗ τ l f 1 (20) twistα (qa . COUNTING CLOSED GEODESICS IN STRATA 29 Consider again the arc ωb in Tb of qb –length of order sb . applying Lemma 4. qb ) i(α. we obtain ∗ lb (19) i(α. Therefore ∗ sb sb lb i(α.17 to ωb and α in qb we have ∗ lb ∗ l b i(α. It remains to show that ωb and α intersect positively. Tb )) ≺ e a √a a l l f r ∗ f ∗a eτ ≺ eτ a ≺p . qb ) − O(1)  i(υαa .6 implies that ∗ b lb  sb and the additive error can be absorbed into the fraction sl b .17. Tb ) ≺ p . Hence. (18) and (21). Therefore. l ExtX (α) eτ By Equation (18).  Lemma 5. Let α ∈ SX and α0 ∈ SZ . we have i(α. ∗ ∗ However. The case when α ∈ SZ can be treated similarly. Tb ) ≺ √ . la f a We now observe that a component of ωb ∩ Fα can intersect υαb at most a bounded ∗ ∗ number of times. `qb (ωb ) ≺ sb and hence `qb (ωa )  lqb (ω). since α is not short in Z. ExtX (α) ∗ eτ (21) twistα (qa . `qb (ωa ) ≥ sb . it has to intersect some boundary curve of Qb . Multiplying the last two inequality. But this is true for every saddle connection of α. Then. sb sb The reason we can ignore the additive errors here is that since α is not short in Z. This proves that ωb intersects ωα essentially positively. by Lemma 4. Tb ) − O(i(α. Then . ωb ) ≺ + O(1) ≺ . `qb (α)  sb and `qb (ωb ) ≺ sb . ωa inter- sects ω at most twice (it can not cross Aaα since its length is less than da ). which means ωb can not twist around α too many times in qb . Tb ) is much smaller than √ .

Assume this leaf intersects the interior of ω3 and makes an acute angle with ω1 inside of the triangle. The fact that edges in Ta intersects edges in Tb essentially positively allows us to find relations between these intersection numbers. then ∗ eτ twistα (X.3. Z) ≺ p . Relations between intersections numbers. and from Theorem 4.1 we have p p (24) ExtX (β0 ) ≤ eτ ExtZ (β0 ). We claim that. we have provided some upper bounds for the intersection number between an edge of Ta and and edge in Tb . γ with α and β with β0 ) there always exits a simple closed curve β0 so that twistα (β0 . multiplying Equations (23). this is equivalent to Equation (2). ςi i(ωi . +1} so that. There is a leaf of the vertical foliation that passes through a vertex of the given triangle before entering it.  5. from Theorem 2. Eventually. ς2 . Z) ≺ ExtZ (α). Z) ExtX (α) ExtZ (α) ≺ eτ . ωb ) = O(1) respectively. for any αb ∈ Sq≥b τ ). we have the relation:   X X (26) ςi i(ωi . But these intersection numbers are not independent. Z) = O(1).14 (this time estimating the length of β0 is X) we have p p (25) i(α. the number of sub-arcs of ωb going from ω1 to ω2 is uniformly bounded. ExtX (α) ExtZ (α0 ) (2) If α = α0 . β0 ) with twistα (X. β0 ). α0 ) twistα0 (X. then ∗ eτ i(α. This is because either the slope of ωb is larger than the . α0 ) ExtZ (α0 ) twistα0 (X. In this section we will describe these relations. On the other hand. Now. Since twistα (β0 . Multiplying the above equation to Equation (22) we obtain part (1) of the lemma. ω2 and ω3 there are sings ς1 .2. β0 ) ExtX (α) ≤ ExtX (β0 ). αb ) = O(1) . For every triangle in Ta with edges ω1 .14 (estimating ExtZ (α)) we have ∗ p p i(α. for every edge ωb in Tb (respectively. since ωb intersects ω1 essentially positively. Z) = O(1) and ∗ p (23) ExtZ (β0 ) ExtZ (α) ≺ i(α. By Corollary 4. COUNTING CLOSED GEODESICS IN STRATA 30 (1) If α0 6= α. our goal is to count the number triangulations that can appear in qb given the triangulation in qa . If α = α0 . There are two kind of relations. from Theorem 4. So far. Lemma 5. i=1.2. ExtX (α) ExtZ (α0 ) Proof. we can replace twistα (X. ς3 ∈ {−1.7.15. (24) and (25) we obtain Equation (22). It is enough to prove that ∗ p (22) twistα (X.3 i=1. β0 ) twistα (X. Z) ≺ p . (replace X with Z. Otherwise. Z) in the above inequality.3 Proof. Z) twistα (X.

Minimality implies that the compliment P is connected.8. we have i(ω1 . Note that B is non-empty since the vertical foliation in qa is not orientable. Let B be the subset of B where the two representatives are negatives of each other (Fig. ωb ) = O(1) respectively. Take a second copy P 0 of P . ωb ) = i(ω3 . Assume that the vertical foliation of qa is not orientable. After orienting the arcs in Ua . Then. we have the relation: ! X X (27) ςω i(ω. Ua ) = O(1). The two vectors are equal up to a multiplication by ±1. +1} so that. consider a saddle connection βα connecting the boundaries of Fα . Now consider a double cover of qa constructed as follows. arcs in Ua generate H1 (S. COUNTING CLOSED GEODESICS IN STRATA 31 slope of ω1 and every time ωb intersects ω1 it has to intersect ω3 next. for every ω̃ ∈ B.  For each α ∈ Sq≥aτ . b. for any αb ∈ Sq≥b τ ). the left side). ω∈B ω∈B Furthermore. The set B = {a. this relation is independent of all the relations in Lemma 5. Let [ Ua = Ta ∪ βα . We now orient edges in Be so that. there is a set B of edges of Ua and for ω ∈ B there is a sign ςω ∈ {−1.7. ωb ) + i(ω2 . e a c a P d c d b e b Figure 4. c}. Σ) where Σ is the set critical points of qa . We can visualize P as a polygon in C with the vertical foliation parallel to the imaginary axis. Note that the signs ς1 = 1. Proof. Glue the edges that were not in B as before and glue the edges in B to the corresponding edge in P 0 . Lemma 5. Σ). Each edge of B has two representatives in the boundary of P . 4). The proof for αb is similar. αb ) = O(1) . or it intersect ω1 a bounded number of times. Hence. . we can think of them as elements of H1 (S. In fact. e P is in the same side of ω̃ (say. Polynomial P . Denote the set of all these edges by B and orient them in some arbitrary way. Choose a minimum number of edges of Ua so that the complement is simply connected. Let Be be the set of lifts of edges in B to this cover. ςω i(ω. for every ωb in Tb (respectively. ς2 = 1 and ς3 = −1 depend only on the triangle and are independent of ωb . ≥τ α∈Sqa We can choose the arcs βα so that twistα (qa . Denote this double cover by q̃a = P ∪ P 0 . ωb ) + O(1).

summing Equation (28) over ω ∈ B. Therefore. Note that q̃a is the unique double cover of qa where q̃a is a square of an abelian differential. The opposite of this is true if ςω = −1. ω̃ 0 )|. b0 . The set Be = {a. That is. We will show that X ςω i(ω. To reiterate. X î(ω̃b . The sum is not zero because ω̃b may start inside P and end in P 0 . ω̃b ) + i(ω̃ 0 . We choose an orientation for ω̃b so it has a positive y–coordinate. ω̃ and ω̃b intersect a uniformly bounded number of times (ω̃b and ω̃ intersect essentially positively). we get X ςω î(ω̃b . ω̃) = O(1). But arcs in Be separate q̃a . ω∈B The proof for the case of a simple closed curve αb ∈ Sq≥b τ is similar. all the intersection points have the same signature. ωb ) = i(ω̃. ω̃∈B e This is because every time ω̃b exits P it intersects the boundary with the opposite signature than when it enters it. Otherwise. ω̃b )  ςω î(ω̃. for every two oriented saddle connections ω̃ and ω̃ 0 in q̃a . Hence. If the absolute value of the slope of ω̃b is larger than that of ω̃ then ω̃b is to the left of ω̃ and hence ω̃ intersects ω̃b with a positive signature. COUNTING CLOSED GEODESICS IN STRATA 32 Let Se be the underlying surface for q̃a and Σ e be the pre-image of Σ. Thus. Polynomials P and P 0 . ωb ) = O(1). Consider ω ∈ B and its lift ω̃. ω̃b ) + î(ω̃ 0 . e e0 0 0 a c a c a a P d P0 d0 c0 c0 0 d d 0 b e b b e0 b0 Figure 5. this is because the number of intersection points that do not have the same sign as ςω is uniformly bounded. either ω̃ intersects ω̃b with a negative signature or a uniformly bounded number of times. . Note that ω̃ has an orientation and hence is identified with vector in C. c0 }. ω̃ 0 ∈ Be are lifts of the same arcs ω ∈ B. ω̃) ≤ 1. If ω̃. b. We define ςω to be +1 if ω̃ has a positive x–coordinate and −1 otherwise. e we let î(  . c. Considering oriented saddle connections as elements of H1 (S. ω∈B Consider an intersection point of ω̃b and ω̃ where ςω = 1.  ) denote the alge- e Σ) braic intersection number. ω̃ 0 ) = |î(ω̃. ω̃b ) . choosing orientation for ω̃ and ω̃ 0 as above we have +   (28) i(ω. a0 . i(ω̃. Let ωb ∈ Tb and ω̃b be a lift of ωb .

We prove the following upper-bound for the size of Bj (Q(σ). α∈SX ExtX (α) Notice that the number of points in Bj (Q(σ). there is a (qa . Also. Given X. Y. Y. X. τ )–regular triangulations Ta and a (qb . Main counting statement This section contains the main combinatorial counting arguments with the goal of proving Theorem 6. Now let. for the associated quadratic differentials qa and qb . Y. τ ) is the same as the number of points in Bj (Q(σ). and  Bj (Q(σ). τ )–regular triangulation Tb that have j common homologically independent saddle connections. X. Recall the definition of B(Q(σ). τ ).1. we note that the relations of the type (26) are also relations in the relative homology with Z2 –coefficients. τ ) ∩ Γ(S) · Y . Y.  B(Q(σ). τ ) is the intersection of the orbit of Y with Bj (Q(σ). COUNTING CLOSED GEODESICS IN STRATA 33 Finally. τ ) to be the set of points Z ∈ T (S) so that. τ ) from §5. Y. τ ) = Bj (Q(σ). X. X. Define Bj (Q(σ). τ ): Theorem 6. τ ). τ ) ∩ Γ(S) · Y . τ ) = B(Q(σ). recall from §2. Consider the stratum Q1 T (σ). Y ∈ T (S) . X. Y. X. X. X. X.1.1. X.  6. this new relation is independent from the previous ones. X. X. But the edges in B are independent in Z2 –relative homology.4 that ∗ Y 1 G(X)  p . Hence. Bj (Q(σ). τ ) ⊂ B(Q(σ). That is.

.

∗ |S |+|S | (h−j)τ .

Bj (Q(σ). Y. τ ). X.

the proof would become significantly more complicated. Remark 6. this is not equivalent to the theorem. First we make a few remarks (1) The theorem still holds if instead of the assumption on the number of common saddle connections. (2) The statement appears to be correct even without the term τ |SX |+|SY | .1. 6. we can still conclude that j ≥ 1. But.2. A marking takes the lengths of the short simple closed curves and the twisting information around short cylinder curves from X and the triangulation and twisting around the non-cylinder short simple closed curves from q. A marking contains a partial triangulation of S. Sketch of the proof Theorem 6. Here is a an outline of our strategy : (1) We define a notion of a marking for the surface S and what it means for a marking to have a bounded length in a Riemann surface X. Fixing a Riemann surface X. when SX ∩ SY contains only homologically trivial simple closed curves. ≺ τ X Y e G(X)G(Y ).1. However. Up to some . dim C+1 where h = 2 . every quadratic differential q where the underlying conformal structure is near X defines a marking that has a bounded length in X. we assume that SX ∩ SY contains at least j homologically independent simple closed curves. a set of short simple closed curves with their lengths and some twisting information.

7 and Lemma 5. whose weighted intersection with ∆0 is less than eτ and so that the intersection patterns satisfy the relations in R. twistα (∆. ∆1 . Given ∆0 and ∆1 and a set of relations R we will define a set MR (∆0 .8. τ ) can then be mapped to a marking in ∆ ∈ MR (∆0 . {E(α)}. Otherwise. Lemma 6. every essential curve in Q intersects T (their free homotopy classes do not have disjoint representatives). if T fills a subsurface S then the set of simple closed curves γ carried by T with `T (α) ≤ B fill the subsurface Q. There is a constant B such that. We denote the set of simple closed curves that are disjoint from T by S c and the remaining curves in S by S n (the set S c is a place holder for large cylinder curves and the set S n is a place holder for non-cylinder  curves or small cylinder curves). R). and • a homotopy class of a partial triangulation T with the vertex set Σ such that the core curve of any annulus in the complement of T is in S. We define a combinatorial length of a simple closed curve γ in Q to be the minimum number of arcs of T that can appear in a representative of γ and we denote it by `T (γ). A partial triangulation T of S with the vertex set Σ is an embedding of a graph to S where vertices are mapped onto Σ and the complementary components are either triangles or annuli. The two notions are related: Lemma 6. T ) = O(1). Recall that a set of curves fill a subsurface Q os S if every essential curve in Q intersects one of these curves. This is similar to assuming that there is a geodesic segment in a stratum Q(σ) staring near X and ending near Y . X. we think of T as representing a free homotopy class of triangulations. again. the main complication is to keep careful track of all the different twisting informations.3. A marking ∆ = ∆ S. An estimate for the number of possible markings ∆0 and ∆1 provides the desired upper-bound for the size of Bj (Q(σ). (that is.2. X.   Definition 6. a relation is a formal linear combination of edges of ∆0 with integer coefficients. As is apparent from the outline. COUNTING CLOSED GEODESICS IN STRATA 34 twisting information. We say a partial triangulation T fills a subsurface Q of S if.4. {E(α)}. Fix a set of point Σ on S. ∆1 . the argument is relatively elementary. (2) Fixing a marking ∆0 .8 provides and upper-bound for the number of elements in MR (∆0 . (3) We then let R be the set of relation of the type described in Lemma 5. τ ). Y. The weights depend on the the length and the twisting information of each short simple closed curve. Y. the expression twistα (∆. We say a marking ∆ = S. ∆1 . This map is finite-to-one except for some twisting information. 6. • for each α ∈ S intersecting T .  ) makes sense). τ ) of all markings ∆ that is a homeomorphic image of ∆1 . T has a bounded length in X if: . Markings on S. We say a curve γ is carried by T if the free homotopy class of γ can be represented by tracing the edges of T . • a length E(α) associated to each simple closed curve α ∈ S. T for S is: • a free homotopy class of oriented curve system S (pairwise disjoint curves) together with a notion of zero twisting for each curve α ∈ S. there are a uniformly bounded number of markings that have bounded length in a given Riemann surface X. τ ) for some marking ∆1 that has bounded length in Y and some marking ∆0 that has both a bounded length and bounded twisting in X. Each Z ∈ Bj (Q(σ). Even though the vertex set is fixed.

The set S and the lengths {E(α)} and the twisting around curves in S c are determined by definition. COUNTING CLOSED GEODESICS IN STRATA 35 (1) S = SX . Here a marking ∆ that has bounded length in X can be obtained as follows: The set S is the set {α. This finishes the proof. X) = O(1). Let M (X.4). The condition (4) for ∆ to have a bounded length in X is a consequence of T being a (q. Proof.19. We continue Example 4. X) = O(τ ). But each of these twisting parameters is bounded by multiple of τ (condition (5) in the definition Definition 6. β} of short curves in X (depicted as blue curves in Fig. q) is depicted in Fig. (2) For α ∈ S. the triangulation T is the (q. Then ∗ |M (X. As it was discussed. there are two thick subsurfaces in the complement of curves α and β (Fig. there is a uniformly bounded number of possibilities for T up to twisting around curves in around curves in S n . Example 6.6. A (q.5. τ )–regular triangulation of (X. (3) For α ∈ S c . 1). 6.  . By Lemma 4. A (q. `X (γ)  `T (γ). 3 4 5 2 6 5 2 6 1 1 4 3 Figure 6. q) described by a gluing of a polygon in R2 . Lemma 6. E(α) = ExtX (α). We say ∆ has bounded length in X with τ –bounded twist if we further have (5) For α ∈ S n . τ )| ≺ τ |SX | . The curves and the triangulation in the marking ∆. twistα (∆. twistα (∆.4 of a surface (X. τ ) be the set of markings ∆ that have a bounded length in X with τ –bounded twist. Figure 7. τ )–regular triangulation. τ )–regular triangulation (depicted as the red trigulation) and E(α) and E(β) are the extremal length of α and β in X respectively. Here Sq≤τ = {α} and Sq≥τ = {β}. ∗ (4) For each simple closed curve γ disjoint from SX . 7). τ )–regular triangulation.

we have ! X X aω i(ω. α∈S c Let Z[U ] be the vector space of formal sums with integer coefficient of edges in U . T } and ∆0 = {S0 . ω ∈ T and ω ∈ T . Note that the partial triangulations in ∆ and ∆ are defined up to homotopy. ∆) E(α) i(α. Then ∗ Y 1 Y 1 |MR (∆. ∆) α. P c (II) For every element aω ω ∈ R and every arc ω ∈ T (respectively. Recall that T and T0 have the same vertex set Σ. τ )| ≺ ehR τ p p . respectively. we have the following bounds on the intersection numbers: ∗ (29) i(ω. α∈S c E(α) α0 ∈S c E0 (α0 ) 0 Proof. T } such that: c (I) ∆ is a homeomorphic image of ∆0 . τ ) to be the set of markings ∆ = {S. let βα be an arc with end point in Σ and disjoint from T that crosses α so that T ∪ βα has bounded twisting around α0 . {E(α)}. aω i(ω. ∆) twistα (∆. {E0 (α0 )}. ω∈U ω c c (III) Given α ∈ S . ω) = O(1). T ) ≺ eτ ∗ p (31) twistα (∆. T0 }. α) twistα (∆. α ∈ S . {E(α)}. α) = O(1) . Let hR = dim(Z[U ]/hRi). ∆) E(α)E(α) ≺ eτ . ω) ≺ eτ ∗ p (30) twistα (∆. T ) ≺ eτ q ∗ (32) i(α. c α∈S . ∆) E(α)E(α) ≺ eτ c and finally if α = α ∈ S c ∩ S we have: q ∗ (33) twistα (∆. Denote [ U =T∪ βα . We give the following upper bound for such markings: Lemma 6. For ∆ ∈ MR (∆. we have E0 (α0 ) = E(α). By above intersection bounds we mean that the homotopy class of two partial trian- gulations have representations with vertex set Σ so that the above bounds hold simultaneously. We define the set MR (∆. ∆0 . ∆0 . Let R be a finite subset of Z[U ] with bounded coefficients. For every α ∈ S c . COUNTING CLOSED GEODESICS IN STRATA 36 Definition 6. Consider the markings ∆ = {S. α ∈ S ).8. and for every α ∈ S that is the image of α0 ∈ S0c . ∆0 .7. Let hRi be the subspace of Z[U ] generated by elements in R. τ ) consider the weighted graphs X W =T+ m(α. ∆) E(α) i(α.

∆) E(α) . ∆0 . COUNTING CLOSED GEODESICS IN STRATA 37 where the weights on the edges of T are 1 and the weight m(α. ∆) ∈ N are defined to be  q  m(α. ∆) = twistα (∆. τ ): n . Define W to be the set of weighted graphs induced by elements of MR (∆.

o W = W .

τ ) . ∆0 . ∆ ∈ MR (∆. .

c ω∈T α∈S Note that in general a weighted graph W is determined by the intersection numbers of its edges with all the edges of U . ∆)). ω) + O dR . ω). ω) ≺ eτ . ω) + mα i(α. Hence. ω) ω∈E is finite to one. ∆) E(α) i(α. To prove the claim. ∆) i(α. ω) = i(ω. ω) = cω i(ω. Equation (29) implies that ∗ (35) i(T . for every ω ∈ W we have ! X X i(υ. • First. . τ )| ≤ p |W|. Consider the set E ⊂ U that forms a basis for the space Z[U ]/hRi. for α ∈ S − S c the Equation (30) implies q ∗ ∗ m(α. ∆0 . there are only finitely many possibilities for the intersection number of W with the other edges of U . We bound the size of I(W) ⊂ Nh by obtaining upper bounds on intersec- tion numbers of W with arcs ω ∈ E. c Also. for α ∈ S . ∆) determines twistα (∆. Consider an edge υ ∈ U . We can find constants cω and dR so that X X υ= cω ω + dR R. the value of m(α. This proves the claim. if ω ∈ T . Hence. there are only finite number of possibilities for i(υ. ω) ≺ twistα (∆. ω∈E R∈R Hence. First. ω∈E R∈R But the constants dR depend only on the set R and otherwise are uniformly bounded. ω) ≺ eτ . The map I records the intersection number with arcs in E. Y 1 (34) |MR (∆. α0 ∈S c E0 (α0 ) 0 We proceed in two steps: Step 1. we claim that the map   I : W → NhR . ∆) only up to √ 1 possibilities (we have used E0 (α0 ) the floor function in defining m(α. c The weighted graph W essentially determines ∆ except that. we need to show that. Step 2. ω). W → i(W . Here we have X X i(W .

by Equation (33) ∗ eτ (39) m(α. COUNTING CLOSED GEODESICS IN STRATA 38 Hence. We will show that Xa and Xb are still near X and Z. we get ∗ Y 1 Y 1 |MR (∆. ∗ (36) i(W . and arc ω ∈ W by Equation (31) we have ∗ ∗ eτ (37) i(βα . ∗ For α ∈ Sq≤âτ . α) ≺ p . Y. qâ ) and (Xb̂ . qb ) so that for any α ∈ Sq≤aτ . Claim: We can choose (Xa . But. (40) twistα (X. qâ ) ≺ twistα (X. ω) twistα (∆.∆) i(α. qa ) and (Xb . ∆0 .16. ∆) ≺ p . E(α) Now from Equations (36). qb ) be the image of (Xb̂ . qa ) to be the image of (Xâ . Assume (Xâ . by Equation (32) we have eτ q ∗ ∗ (38) m(α. we get: ∗ ∗ Y eτ Y |W| ≺ |I(W)| ≺ p × eτ α∈S c E(α) ω∈T ∩E βα ∈E Y 1 ≤ e|E|τ p . as before. ExtX (α) Hence. We need to be a bit careful. (37). There may be many choices for these quadratic differentials. qb̂ ) under the same homeomorphism.o∆) i(α. Xâ ) + twistα (Xâ . using the triangle inequality and Theorem 2. τ )| ≺ ehR τ p p . ExtX (α) . ω) ≺ eτ .1. if α = α ∈ S c ∩ S . assume that they do not satisfy Equation (40). X. α∈S c E(α) α0 ∈S c E0 (α0 ) 0 which is as we claimed. ∆) E(α) i(α. We define (Xa . ∗ 1 twistα (qâ . qâ ) under an appropriate number of Dehn twists around curves in Sq≤âτ to ensure (40) and let (Xb . Let Z ∈ Bj (Q(σ). α∈S c E(α) Now.  Proof of Theorem 6. by Lemma 4. • For arc βα ∈ U where α ∈ S c . qa ) and (Xb . qâ ) = O = O(τ ). qb ) be the initial and the terminal quadratic differentials for the Teichmüller geodesic in Q(σ) starting near X and finishing near Z ∈ Γ · Y . qa ) = O(τ ). βα ) ≺ p . E(α) c Finally. E(α) c And for α ∈ S − S c . if ExtX1 (α) ≺ τ . α) ≺ twistα (∆. Xqa ) ≺ . applying Equation (34). qb̂ ) are some choice of initial and terminal points with associated regular triangulations Tâ and Tb̂ that have j common saddle connection.2   + 1 (41) twistα (X. Proof of claim. τ ) and let (Xa . ω) ≺ i(α. (38) and (39).

∆0 . Let ∆0 = g−1 (∆). using the fact that ∆ is bounded in Z and Equation (43). note that S c = Sq≥aτ and S n = Sq≤aτ . ≤τ • If α ∈ Sqa then set the twisting around α in ∆ so that twistα (∆.8 provides an upper-bound for the size of the set MR (∆. we have ∆ ∈ MR (∆. Since α is a non-cylinder curve. (Xa . ≤τ α∈Sqâ Hence. τ )–regular triangulation Tb . τ ). We can similarly associate a marking ∆ to the pair (Xb . Also. the images Ta and Tb of Tâ and Tb̂ are still regular triangulations and have j arcs in common. Instead. Xa )  nα ExtX (α) ≺ 1.16. To the pair (Xa . Hence. ≤τ α∈Sqâ and ∗ ∗ X dT (Yb̂ .8 in case quadratic differentials in Q(σ) are not orientable) and the j edges in T that are present in the (qb . T ) = O(1). τ ). Y Xa = Dαnα Xq0 . The result is a marking that has bounded length in X and (by Equation (40)) has τ –bounded twist in X. By. • If α ∈ Sq≥aτ then set the twisting around α in ∆ so that twistα (∆. ≤τ α∈Sqâ ∗ where Dα is a Dehn twist around α and nα ≺ ExtX1 (α) . T } as follows: • Let S be the set of short curve in X and set E(α) = ExtX (α). we have 1 ∗ 1 (42)  . • Let T be the (qa . Then ∆0 in bounded in Y . for τ large enough. qa ) and (Xb .  For the rest of the proof. Also. qa ) we associate the marking ∆ = {S. qb ). qb ) are as desired.7.6) and there are finitely many choices of for the homeo- morphism type of ∆0 . Xb )  nα ExtZ (α) ≺ 1. similar to Equa- tion (41). we assume Equation (40) holds.2 ∗ ∗ X dT (Xâ . Also. . qb ) ≺ . ∆0 . X) = O(1). Now. this is because the inequality (40) does not necessarily hold for Z and qb . ExtZ (α) Assume Z = g(Y ). {E(α)}. (40) is holds and no modification is required. COUNTING CLOSED GEODESICS IN STRATA 39 Therefore. Theorem 2. Lemma 6. let R be the elements in Z[U ] coming from Lemma 5. we have ∗ 1 (43) twistα (Z. assume ExtX1 (α) ≥ τ . Taking this Tb is the partial triangulation in ∆. for g ∈ Γ(S). ExtX1 (α) changes at t most linearly with time (Equation (7)). That is. ExtX (α) ExtZ (α) Again by Lemma 4. Here we can only conclude that ∆ is bounded in Z (not with bounded twist). τ )–regular triangulation Ta which has j edges in common with the triangulation Tb . Also. The number of possible choices for ∆ is O(τ |SX | ) (Lemma 6. the number of Dehn twists nα around α that needs to be applied to qâ to ensure Equation (40) is at most O(1/ ExtX (α)). (and Lemma 5.

That is. ∆0 . These estimates imply Theorem 1. 1 ∗ 1 1 ≺p p .6. The proof is finished after checking that hR = (h − j). X. Then. and α ∈ Sq≤aτ ∩ Sq≤b τ = S n ∩ S0n . The main idea. Y.3).1. τ ) is comparable to the product of the following: the number of choices for ∆. Moreover. ∆0 . τ ). τ ) to the union of sets of markings MR (∆. α∈S E(α) α0 ∈S0 E0 (α0 ) The Q last line1 follows from the previous line because. X.5. as in Equation (42). every closed geodesic in C can be approximated by a path along the net points. COUNTING CLOSED GEODESICS IN STRATA 40 similar Q to Lemma 6. X. the number of choices for the homeomorphism class of ∆0 . the maximum multiplicity of the association Z 7→ ∆ and the size of MR (∆. this is because. by fixing an appropriate net N e in T (S). τ )| × O(1) × × |MR (∆. which shows that the flow (or more precisely an associated random walk) is biased toward a compact part of the space. λi+1 )}i≥0 (see §3. τ )| ≺ |M (X. See Lemma 7.4 for the precise formulation. Lemma 5.5) tend not to have short saddle connections. we have defined a map from Bj (Q(σ). the dimension of Z[U ]/hRi is exactly j less than h = dim2C+1 . . Y.1 and Theorem 1.  7. ∗ Y 1 |Bj (Q(σ). The fact that the j arcs we have fixed in Ta are homologically independent implies that these arcs and the other relations in homology are independendant in Z[U ]. roughly speaking. we consider the random walk {λi }i≥0 on the points in N and apply Theorem 6. In fact. This is true because all the relations in Lemma 5. where ∆ is bounded X with τ –bounded twist and ∆0 is bounded in Y . we show that quadratic differentials {q(λi . τ )| ≤τ ExtZ (α) α∈Sqb ∗ Y 1 Y 1 Y 1 ≺ τ |SX | ehR τ p p ExtZ (α) E(α) α0 ∈S c E0 (α0 ) ≤τ α∈Sqb α∈S c 0 ∗ Y 1 Y 1 ≺ τ |SX |+|SY | ehR τ p p . But this is accounted for in the definition of h (see §3. We discretize the projection π(Q(σ)) ⊂ T (S). ∆0 . The size of Bj (Q(σ). Y. b To summarize.8 is used only when ς = −1. Σ). is to prove an inequality. which is due to Margulis. ExtZ (α) ExtZ (α) ExtX (α) That is. τ ). The map is not one-to-one but we have a bound on the multiplicity. Hence. each term can either be counted in the power of τ in the beginning of last line or it can be divided into a term in each of the last two products. Geodesics in the thin part of moduli space In this section we prove Theorem 1. Consider the stratum Q(σ).7 are also relations in H1 (S. for every term in the product 1 α∈Sq ≤τ b ExtZ (α) we either have ExtZ (α) = O(τ ) or.  we can conclude that the association Z 7→ ∆ is at most 1 O α∈Sq ≤τ ExtZ (α) –to-one.1 to show that the projection e of this random walk in M(S) is biased towards the compact subset of M(S).

) to denote the maximum number of homologically inde- pendent disjoint saddle connections in Ωq (). Define s(q. q) ∈ Q1 T (S). recall the set of short saddle connections Ωq () (Def- inition 4. Short saddle connections and simple closed curves. For a quadratic differential (X. Given the tuple σ. define n . COUNTING CLOSED GEODESICS IN STRATA 41 7.8).1.

q) ∈ Q(σ) . (σ) = (X. o Qj.

. s(q. ) ≥ j ⊂ Q1 T (S).

. X. X. . recall the definition of B(Q(σ). X. Proof. for each i. X. Also.13. X. Y. we have s(qa . It is enough to let 2 (τ ) = e−2τ 1 . ). Namely. τ ). X. τ ) to be the set Z ∈ T (S) so that • Z ∈ B(Q(σ). τ ) from §5. . . τ ). Roughly speaking. Assume. . . qa and qb are the associated quadratic differentials and (b − a) < τ . . Let ω1 . with fix σ and denote Qj. X. τ ). Y. only. Z ∈ B(Qj. we have (44) B(Qj. τ ) as in §6. X. define B(Qj. Then. . Lemma 7. For the rest of this section. For any τ > 0.1 and Bj (Q(σ). We would like to refine the definition of Bj (Q(σ). (σ) simply by Qj. . X. τ ) ⊂ Bj (Q(σ). there is 2 (τ ) < 1 (τ ) such that for  < 2 (τ ). by Equation (8). τ ) • for the associated quadratic differential qa . ωj be disjoint homologically independent saddle connections counted in s(qa . and any X. ) ≥ j.1. and (45) B(Qj. τ ) from §6. τ ). any integer j ≥ 0. . we are interested in a ball of radius τ centered at X that is allowed to move in the direction Qj. Recall the choice of 1 (τ ) from Lemma 4. X. τ ) ⊂ Bj (Q(σ). Y. X. One can similarly define B(Qj. . Y ∈ T (S).

.

.

1 1 .

.

Extq (ωi ) − Extq (ωi ) .

≤ τ. .

.

2c)–separated net N ⊂ M(S) we mean a set of points in M(S) so that: • the Teichmüller distance between any two net points in N is at least c. and (c. The proof of Equa- tion (45) is similar.2.13). . τ ) = p(B(X. τ )–regular triangulation Ta and a (qb . τ )| ≺ τ 3g−3+p . and • any point in M(S) is within distance 2c of a point in N . 2c) net N e as above. it is easy to check (see Lemma 3 in [EM2]): Lemma 7. There exists a constant c0 > 0 such that for any c > c0 . Then. Hence. we have ∗ (46) |N (X. τ ). That is. by the definition Z ∈ Bj (Q(σ). Thus.  7.2. τ )) ∩ N . ωi ∈ Ωqb (1 ).1 the extremal length of any short curve α containing ωi changes by at most a factor of at most e2τ . τ )–regular triangulations Tb (Lemma 4. a b and by Theorem 2. the set {ωi } can be extended to both a (qa . X. The arcs ωi are still disjoint and homologically independent in qb . Let N (X. By a (c. Choosing a net.

. X. . and  < 2 (τ ) we have ∗ Aτj. τ ) = B(Qj. τ ) = B(Q(σ). f : T (S) → R. X. for X. τ ) ∩ N e. . Proof. τ ) ∩ N e. Y. G (X) ≺ τ m e−jτ G(X). consider the average function Aτj. By Lemma 2. yk and let Yi ∈ T (S) be a pre- image of yi . X. . . . 7. X. X. i = 1. For a real-valued function f : M(S) → R. N e (Q(σ). . . The main inequality. τ ) as y1 . τ ) ∩ N e. . τ ) ∩ N e.  (47) where m depend only on the topology of S.τ ) Here. X. Y. . τ ).3. and N e (Qj. . X. with N e (  ). Z∈N e (Qj.  defined by X Aτj.3. X. k. 2 Our main tool is the following (2 (τ ) is as in Lemma 7.X. Given τ > 0. Yi . X. Y. Enumerate the elements of N (X. as before dim C + 1 h= . .1): Proposition 7. . Y ∈ T (S). Y. B(  ). τ ) = B(Q(σ). COUNTING CLOSED GEODESICS IN STRATA 42 Let N e = p−1 (N ). We denote the intersection of a ball in Teichmüller space.3. N e (Qj. every net point in Z ∈ B(Qj. That is. Yi . τ ) is near at most G(Yi ) points in B(Qj. . e by N N e (Q(σ). That is. τ ) = B(Qj. f (X) = e−hτ  f (Z). . X. . Ne .

.

.

.

∗ .

(48) .

τ ). X.N (Qj. . Yi .

G(Yi )2  .

τ ). X. Yi . .B(Qj.

.

e .

.

 . Hence.X. G (X) = e−hτ  G(Z) Z∈N e (Qj.Yi . .τ ) k X X = e−hτ G(Yi ) i=1 Z∈N e (Qj.τ ) .X. we have X Aτj.

.

k .

.

Y . τ ).B(Q . X.

.

 i (Equation (48))  e−hτ i=1 G(Yi ) k ∗ X ( Theorem 6. Here.  . ∗ X j. m = (9g − 9 + 3p) ≥ |SX | + |SY | + (3g − 3 + p).1 and (45)) ≺ e−hτ τ |SX |+|SY | e(h−j)τ G(X) i=1 −jτ m (Equation (46)) ≤e τ G(X).

λk−1 ) ≤ τ . we have dT (λk .τ (Qj. λk−1 . Suppose R  τ and let n be the integer part of R/τ . Let Pτ (X. R) denote the set of all trajectories λ ∈ Pτ (X. n} → N e ⊂ T (S) such that. For j ∈ N. let Pθ. By a trajectory of the random walk we mean a map λ : {0. X) = O(1). R) so that. τ ). COUNTING CLOSED GEODESICS IN STRATA 43 Trajectories of the random walk. X. • . R) denote the set of all trajectories for which dT (λ0 . . for all 0 < k ≤ n. • for 1 ≤ k ≤ n λk ∈ N e (Q(σ). where λk = λ(k).

 .

τ ) . e .

.

. ≥ θ · n.

λk ∈ B(Qj. λk−1 . . N . k | 1 ≤ k ≤ n.

R) = Pθ. Finally. X.τ (Qj. . .4.1 and Proposition 7.3. Given X. p(λn ) = O(1). . X. Y ∈ T . R) denote the set of all trajectories λ ∈ Pθ.τ (Qj. R) such that  dT p(Y ). Let 2 (τ ) be as in Lemma 7. 0 ≤ θ ≤ 1 and  < 2 (τ0 ) we have . X. We say that a trajectory is almost closed in the quotient if  dT p(λ0 ).τ (Qj. Lemma 7. . For any δ0 > 0 there is τ0 > 0 so that for τ > τ0 .τ (Qj. R) denote the subset of these trajec- tories starting from X which are almost closed in the quotient. Y. p(λn ) = O(1). let P bθ. let Pθ. X. X.

.

∗ (h−jθ+δ )R G(X) (49) .

Y. .Pθ. R).τ (Qj. X.

. G(Y ) In particular. ≺ e 0 .

.

∗ (h−jθ+δ )R (50) .

R).τ (Qj. .Pbθ. X.

Proof. Define . ≺ e 0 .

 .

.

qk (λ) = .

i .

τ ) . λi−1 . . 1 ≤ i ≤ k. λi ∈ N e (Qj.

.

..

 . R) but truncated after k = r/τ steps. λ∈Pθ. . Y.r) Also. r) be the set of trajectories obtained from a trajectory λ ∈ Pbθ.τ (Qj. let Pbθ. R)| ≺ .τ (Qj. . Y.X. Define X Vτ (R. For 0 < r = kτ < R. λ∈Pθ. Y.Y.τ (Qj. . X.X. R. . G(Y )ejθR .τ (Qj. Therefore. X. let R = nτ . ∗ Vτ (R) (51) |Pθ. q(λ) = qn (λ) and X Vτ (R) = G(λn )ejq(λ)τ .R. .R) ∗ Note that G(Y )  G(λn ) and q(λ)τ ≥ θR. r) = G(λk )ej qk (λ)τ .τ (Qj.Y. This keeps track of the number of steps in the trajectory λ (amount the first k steps) that can be approximated by a segment in Qj. X.

R. We choose τ large enough so that m log(τ ) + log(C) < δ0 . R) may be much larger than Nθ (Cj. Lemma 7.τ (Qj. R). X. and m ∈ N are universal constants. for τ > τ1 X ∈ T (S) and any sufficiently large R (depending only on δ1 . Now iterating (52) n = R/τ times we get m(log(τ )+log(C)) (53) Vτ (R) ≤ (C τ )mn G(X) ehnτ = G(X)e(h+ τ )R . . there is τ1 so that. .1).Y. s(qt .r+τ )  X X ≤  G(λk+1 )ej(qk (λ)+1)τ + λ∈Pθ. Y. Note that the right hand sides of the above two equations are the same.R. qt ) spends θ proportion of time in Qj. . For any δ1 > 0. X. λk+1 6∈N e (Qj.τ ) The two summands inside of the parenthesis are similar to the average defined above.X. R) be the number of conjugacy classes mapping classes associated to closed geodesics g in C of length at most R which pass within a uniformly bounded distance of the point x and so that for at least θ fraction of the points (xt . G)(λk ) ≺ ej(qk (λ)+1)τ ehτ τ m e−jτ G(λk ).τ )  X + G(λk+1 )ejqk (λ)τ  . Hence. the first term is less than (up to a multiplicative error) ∗ ej(qk (λ)+1)τ ehτ (Aτj. Y.X. τ The lemma follows from Equation (51) and Equation (53).R. for x ∈ M(S). δ) ≥ j (see §7. up to a multiplicative error) ∗ ejqk (λ)τ ehτ (Aτ.5.  Let Nθ (Qj. Similarly. . τ ) then qk+1 (λ) = qk (λ)+1 and qk+1 (λ) = qk (λ) otherwise.λk . Hence. X.0 G)(λk ) ≺ ejqk (λ)τ ehτ τ m G(λk ).λk .r) λk+1 ∈N e (Qj. where C > 0.τ (Qj. .Y. . . x. . λk .X. Nθ (Qj. X Vτ (R. r + τ ) = G(λk+1 )ejqk+1 (λ)τ λ∈Pθ. COUNTING CLOSED GEODESICS IN STRATA 44 If λk+1 ∈ N e (Qj. r + τ ) ≺ τ m ehτ ejqk (λ)τ G(λk ) λ∈Pτ (C. let Nθ (Cj. Using Equation (47). r). ∗ X Vτ (R.r) m hτ (52) =τ e Vτ (R. qt ) ∈ g. . for x = p(X). τ ) we have . and the second term is less than (again. As we shall see in the proof of the lemma below. x. . X. R) be the number points Z ∈ B(Q(σ). R) (see §6 for definition) so that associated geodesic (Xt .

∗ .

.

p(X). (1 − δ1 )R) ≺ . (54) Nθ (Cj. .

τ (Qj. . X. R).Pθ.

..

.

COUNTING CLOSED GEODESICS IN STRATA 45 and .

∗ .

.

X. (55) Nθ (Qj. (1 − δ1 )R) ≺ . Y. .

 .τ (Qj. X. Y. R).Pθ.

. G(Y )2 .

Proof. Recall the definition of  .

IX = g ∈ Γ(S) .

g · X) = O(1) . . (1 − δ1 )R). for each of these we consider all the possible lifts which pass near X. We get: ∗ . x. (This replacement is the cause of the δ1 R error). If the original geodesic g has length at most (1 − δ1 )R and has s(qt . X) = O(1).3. We can mark points distance τ apart on G. If two geodesic segments map to the same trajectory. We now consider all possible geodesics contributing to Nθ (Cj. Then there are approximately |IX | lifts of [g] to Tg which start within a bounded distance of X. thus g−1 2 g 1 ∈ I X . then the segments fellow travel within O(1) of each other. . from Lemma 2. Each lift G is a geodesic segment of length equal to the length of g. Consider a closed geodesic g in C which intersects a uniformly bounded neighborhood of x = p(X). and then for each lift consider the associated random walk trajectory. dT (X. R). then the resulting trajectory belongs to Pθ. and replace these points by the nearest net points in N e . In particular if g1 and g2 are the pseudo-Anosov elements corresponding to the two geodesics. Let [g] denote the corresponding conjugacy class in Γ(S). X. This gives a map Ψ from lifts of geodesics to trajectories. then dT (g−1 2 g1 X. δ) ≥ j for θ fraction of its points.τ (Qj.

.

 . |IX | Nθ (Cj. (1 − δ1 )R) ≺ |IX | . x.

 . R).Pθ. X.τ (Qj.

and the factor of |IX | on the right is the maximum possible number of times a given random walk trajectory can occur as a result of this process. An argument similar to the proof of the first part implies Equation (55). The factor of |IX | on the left hand side is due to the fact that we are considering all possible lifts which pass near X.6. Thus Theorem 1. ∗ (56) Nθ (Cj. Suppose g is a closed geodesic of length at most R on M(S). Note that by ∗ Lemma 2.  . for R large enough.. Choose δ0 .1) |I(Y )| ≺ G(Y )2 . x∈N where N is the net chosen above. the number of relevant points in the net is at most polynomial in R. Thus. . In view of Lemma 7.5.  We need he following lemma which is due to Veech [Ve]. R) ≤ Nθ (Qj. Theorem 6. x. R) ≺ e(h−jθ+2δ/3)R . this polynomial is less than eδR/3 .5 follows. Let δ > 0. Finally X Nθ (Cj. the factors of |IX | cancel. x. and the lemma follows. τ1 } and let R be large enough so that Equations (50) and (54) hold. Lemma 7. R).2.3 (See also. Then for any x ∈ g. any X so that p(X) = x and every simple closed curve α ∗ ExtX (α)  e−(6g−4+2p)R . . We get.6 and Lemma 7. However. Now choose τ ≥ max{τ0 . Proof of Theorem 1. δ1 ≤ δ/3.

Let g be a closed geodesic in C \ K. We can use the argument applied in the proof of Theo- rem 1. Hodge norm. Hence the number of disjoint homologically independent saddle connections in Ωq () is at least one. In this section. we consider the canonical (ramified) double cover φ : X̃ → X such that φ∗ (q) = φ2 .7) follows from Lemma 7. We choose K so that Lemma 4. is to pass to the possibly ramified double cover on which the foliation defined by q is orientable.0 . Let π : H1 T (S) → T (S) and p : H1 T (S) → H1 M(S) be natural maps as in §5. As in the proof of Theorem 1. A standard con- struction. Choose a net satisfying Lemma 7. The Hodge Norm and the Hodge Distance.8 for the explicit construction. By taking K large enough we can assure that every quadratic differential along g has an arbitrary short saddle connection. τ1 } and let R be large enough so that Equations (49) and (55) hold.2. Let k  kH. 8. This yields a surface X̃ with an abelian differential φ. COUNTING CLOSED GEODESICS IN STRATA 46 Proof of Theorem 1. we need a similar property for compact subsets of the spaces Q1 M(σ) of quadratic differentials. =(φ) ∈ H 1 (X.0 . R) be forms obtained by the real part and the imaginary part of the holonomy.2.5 below.  8. there may be a curve that has a very small extremal length in X̃.t denote the Hodge norm on the surface Xt = π(gt φ). q). The theorem now follows from Theorem 1. More precisely. φ) in H1 T (S).1 gives a partial hyperbolicity property of the geodesic flow on spaces of abelian differentials. Ωq () is non-empty for  ≤ 2 (τ ). Let H1 T (S) be the bundle of area one abelian differentials over T (S).9 implies that any such quadratic differential (x. Quadratic and abelian differentials.1. Equation (1. then we have kλkH. In our application. gt φ] starts and ends in p−1 (K) and spends at least half the time in p−1 (K). we use the Hodge norm [Fo] to show that in any compact subset of C the geodesic flow is uniformly hyperbolic: see [ABEM] and Remark 8. (See the proof of Lemma 5. the segment [φ. R) and any t ≥ 0. we briefly treat the case when q ∈ QM(S) is not the global square of an abelian differential.2 and Lemma 7.6.1. Then choose τ ≥ max{τ0 . kλkH. The following fundamental result is due to Forni [Fo.7. 8. let <(φ). in addition. Let 0 < δ0 . .5. Also. Theorem 8.5.  Proof of Theorem 1.1. Here. R) for j = 1 and θ = 1.t ≤ et kλkH. However. for each abelian differential φ. g is counted in Nθ (Cj. That is.1. even if p(X) belongs to a compact subset of M(S). given X ∈ T (S) and q a quadratic differential on X. for some compact subset K of H1 M(S). We also denote by gt the geodesic flow on H1 T (S) (where we square an abelian differential to get a quadratic differential).t ≤ e(1−α)t kλkH. λ∧<(φ) = λ∧=(φ) = 0 and. where X ∈ T (S) and φ is an abelian differential on X. where α > 0 depends only on K.5. §2]: Theorem 8. If.) The set of critical values of φ coincides with the set of zeros of q with odd degree. This may occur since the flat structure defined by q may . Fix a point (X. For any λ ∈ H 1 (X. δ1 ≤ δ/3.

if q does not have any short saddle connection. To simplify the notation. Let X̃ be as before and Σ̃ be the pre-image of Σ. R) denote the natural map. R) as follows: . Let (X. That is. We have. Σ̃. On X̃. q̃ has a canonical square root which we denote by φ.3. `min (φ) ≥ `min (q). 8. q) ∈ Q1 T (σ) and let Σ be the set of singularities of q. Σ̃. let X̃ = X. Σ̃ = Σ. COUNTING CLOSED GEODESICS IN STRATA 47 have an arbitrarily short saddle connection connecting distinct zeroes. if q is a square of an abelian differential. R) → H 1 (X̃. The Hodge norm on relative cohomology. then φ also does not have any short saddle connections. We define a norm k  k on the relative cohomology group H 1 (X̃. Such a saddle connection lifts to a very short loop in the double cover. Let `min (q) denote the length of the shortest saddle connection in the flat metric defined by q. Let j : H 1 (X̃.

Z .

X .

.

(57) kλk = kj(λ)kH + (λ − h).

. .

.

.

0 (p.p )∈Σ×Σ .

p0 . γ p.

p(qt ) ∈ K and that the geodesic segment [q0 . quadratic differentials in K have no short saddle connections. Proof of Theorem 8. Since j(λ) and h represent the same class in H 1 (X̃. h is the harmonic representative of the cohomology class j(λ) and γp. the Equation (57) does not depend on the choice of γp. Let K be a compact subset Q1 M(σ). qt ] spends at least half the time in p−1 (K). We reproduce the proof here for the convenience of the reader.1: Theorem 8.2. Let qt . Xt and φt be defined as usual and let k  kt denote the norm (57) on the surface X̃t = π(gt φ). Then we have kλkt ≤ e(1−ᾱ)t kλk0 . R) with j(λ) ∧ <(φ) = j(λ) ∧ =(φ) = 0. any harmonic ψ ∈ H 1 (X̃u . for u ∈ [0. Suppose λ ∈ H 1 (X̃. R). R) and any arc γ on X̃u with end points in Σ̃. Hence. p(qu ) ∈ K implies that X̃u is thick (has no curves with short extremal lengths). Then there is t0 > 0 so that for t > t0 the following holds. We have the following analogue of Theorem 8. Since K is compact. where ᾱ > 0 depends only on K. Therefore. . This theorem is essentially in [AF] (Lemma 4.p0 is any path connecting the zeroes p and p0 .2. where k  kH denotes the Hodge norm on H 1 (X̃. Suppose p(q0 ).4).p0 . Σ̃. t]. there exist a constant cK such that for any u with p(qu ) ∈ K. R).

Z .

.

.

 (58) .

ψ .

≤ cK kψkH.u 1 + `u (γ) . .

.

0. Under the assumptions of Theorem 8. γ where `u (γ) is the length of γ in flat metric associated to φu . Fix p. Since X̃0 is thick.1t. there exists s ∈ [0. there exists a path γ0 connecting p and .9t] such that p(qs ) ∈ K.2. p0 ∈ Σ.

Let ψ = j(λ). `0 (γs ) = O(es ) and `s (γt ) = O(et−s ). Similarly. Suppose λ ∈ H 1 (X̃. Σ̃. since X̃s and X̃t are thick there are paths γs and γt connecting p and p0 such that `s (γs ) = O(1). COUNTING CLOSED GEODESICS IN STRATA 48 p0 with `0 (γ0 ) = O(1). `t (γt ) = O(1). Then. We have . For 0 ≤ u ≤ t. R) with j(λ) ∧ <(φ) = j(λ) ∧ =(φ) = 0. let ψu denote the harmonic representative of the cohomology class ψ on X̃u .

.

X .

.

Z .

kλkt − kλk0 ≤ kψkH.t − kψkH.0 + ψt − ψ0 .

.

.

p.p0 ∈Σ×Σ .

p0 . γp.

.

Z .

X .

.

.

(59) ≤ e(1−α)t kψkH.0 + ψt − ψ0 .

. .

.

p ∈Σ×Σ . 0 p.

p0 . γp.

1. where we have used Theorem 8. Since the integral in Equation (59) is independent of the choice of γp. by Equation (58).p0 . . we use γp. Then.p0 = γs .

Z .

∗ ψ0 .

.

.0 (1 + `0 (γs )) ≺ cK kψkH.0 es . ≤ cK kψkH.

.

(60) .

.

. γs Also.

Z .

.

Z Z .

.

.

.

.

.

.

ψt .

.

= .

.

ψt + ψt .

.

γs γ −γ γt .

Z s t Z .

.

.

=.

.

ψs + ψt .

.

γ −γ γ .

Z s t.

.

Z t .

.

Z .

.

.

.

.

.

.

≤.

.

ψs .

+ .

.

.

ψs .

+ .

.

.

ψt .

.

s + kψkH. Then. we have . R). where to pass from the first line to the second we used the fact that ψs and ψt represent the same cohomology class in H 1 (X̃.s e + kψkH. using Equation (60). and in the last line we used Equation (58) to estimate each term.t . γs γt γt ∗   t−s ≺ cK kψkH.

Z .

  .

∗ t−s + kψkH.t + kψkH.0 es .

.

.

ψ t − ψ 0 .

s + kψkH.s e . ≺ cK kψkH.

Assuming t is large enough.1t.1α)t kψkH. γs   ≤ cK e(1−α)s + e(1−α)s+t−s + e(1−α)t + es kψkH. Substituting into Equation (59) we get kλkt − kλk0 ≤ cK e(1−0. 0.0 ≤ cK e(1−0.1α)t kλk0 .1α.1 and in the last line we use the fact that s ∈ [0. The theorem then holds for ᾱ ≤ (0.0 .1α − α0 ).0 ∗ ≺ cK e(1−0. where in the second line we used Theorem 8. we can assume that the multiplicative error is less than eα0 t for som α0 ≤ 0.  .9t].1α)t kψkH.

q0 ) = {(X.  (62) Then dH (gt q. We also consider the foliation F u whose leaves are defined by [ αu (q) = gt αuu (q) t∈R s and F whose leaves are defined by [ αs (q) = gt αss (q). for ψ ∈ W− (or ψ ∈ W + ). q 0 ) for any q. Lemma 8. φ0 ) to be the infimum of `(γ) where γ varies over paths connecting φ and φ0 and staying in the leaf of F ss ⊂ Q(σ). Note that the foliations F ss .  We now show that the above condition holds whenever the projections of gt q and gt q 0 to C are also close. s ∈ [0. Proof. This follows from Theorem 8. q) | η + (q) = const . R). then we define the Hodge length `(γ) of γ as 0 kγ (t)k dt. q 0 on the same leaf of F ss in Q(σ). (2) F uu . Let K be a compact subset of C. r] into some leaf of F ss . a leaf of F ss is given by αss (X0 . Let gt be the Teichmüller flow on Q1 M(σ). COUNTING CLOSED GEODESICS IN STRATA 49 8. in partic- ular.2). (X 0 . F uu are invariant under both gt and Γ(S). q).2 and Equation (61). q0 ) = {(X. We make the same definition if φ and φ0 are on the same leaf of F uu . q) ∈ Q(σ) | η − (q) = η − (q0 )}. To each quadratic differential q.4. we can locally identify a leaf of F ss (or F uu ) with a subspace W− (or W + ) of H 1 (X. See also Lemma 5. In other words. we associate its imaginary and real measured foliations η − (q). q 0 ). The Hodge Distance. then we define dH (φ. See §1 and §2 of [Fo] for more details. we can define dH (q. If γ is a mapR r from [0. Suppose t > t0 is such that for all q 00 ∈ γ. where c depend only on K. q) ∈ Q(σ) | η + (q) = η + (q0 )}. for (X0 . and a leaf of F uu is given by αuu (X0 . t∈R If C is a subset of moduli space of abelian differentials. whose leaves are sets of the form (X. q) | η − (q) = const . q0 ) ∈ Q(σ). Finally: 0 • If two abelian differentials φ and φ0 belong to the same leaf of F ss . Suppose (X. we have (61) j(ψ) ∧ =(ψ) = 0 and j(ψ) ∧ <(ψ) = 0. Σ. Let γ be a Hodge length minimizing path connecting q to q 0 . where k  k is the Hodge norm. and η + (q).3. t] | gs q 00 ∈ p−1 (K) ≥ t/2. gt q 0 ) ≤ e−c t dH (q. The flow gt admits the following foliations:  (1) F ss . . they descend to the moduli space Q1 M(σ). • By taking a ramified double cover (see §8. q 0 ) ∈ p−1 (K) are in the same leaf of F ss . In fact.4 of [EM2]. whose leaves are sets of the form (X.

t] | p(gs q 0 ) ∈ K0 ≥ t/2. Let (X. Then there is a larger compact subset K0 ⊂ C and a covering of K with a finite family of open sets U so that the following holds. for some s ∈ [0. Then. COUNTING CLOSED GEODESICS IN STRATA 50 Lemma 8. gt (q 0 ) ∈ U2 . and let (X1 . (X 0 . for contradiction that `qs (ω) > ρ0 `qs0 (ω). 2. i = 1. (X2 . q 0 ) ∈ U1 and t > 0 be such that gt (q). Let U1 . we move forward in time and argue the same way. . First. 9.4. 2. q2 ) ∈ U . we prove Theorem 1. We essentially follow the work of Margulis [Mar]. Let U be connected open sets so that p(U ) ∈ U. Then for any saddle connection ω. 2 2 Which contradicts Equation (65). gt (q 0 ) ∈ U2 .  (64) Proof. U2 ⊂ Q(σ) be connected open sets so that p(Ui ) ∈ U. U2 ⊂ Q(σ) be connected open sets so that p(Ui ) ∈ U.2. We can find an open cover U of K so that the following holds. q1 ). Assume. In case ω is mostly horizontal. and let K0 be the compact subset of C consisting of quadratic differentials where the length of every saddle connection is larger than 0 = /ρ0 . We have essentially shown that under the assumption Equation (63) we have exponential contraction along the foliation F ss (and similarly exponential expansion along the foliation F uu ). s ∈ [0. (X 0 . Further. we have 1 (65) `q (ω) ≤ `q2 (ω) ≤ ρ `q1 (ω). 2 s Then `q (ω) ≥ =(holq (ω)) = es =(holqs (ω)) 1 > es `qs (ω) 2 1 1 > es ρ0 `qs0 (ω) ≥ ρ0 `q0 (ω).  Remark 8. ρ 1 Let U1 . Let (X. Let K be a compact subset of C. assume that  (63) s ∈ [0. t]. q 0 ) ∈ U1 and t > 0 be such that gt (q). This proves the claim. q). t] | p(gs q) ∈ K ≥ t/2. We first claim that (65) is true for quadratic differentials qs = gs (q) and qs0 = gs (q 0 ) as well for a larger constant ρ0 = 2ρ. Assume ω is mostly vertical in qs .5. Then (64) follows from the above length comparison. 1 =(holqs (ω)) > `q (ω). We only outline the arguments here since they are well known a more detailed version is already present in [H3]. Let ρ > 0.2 In this section. q). i = 1. Now let  be such that the length of any saddle connection in q ∈ K is larger than . That is. we need a closing lemma. Outline of the proof of Theorem 1.

In other words. Outline of the proof of Lemma 9. (III) g0 passes through U 0 . q2 ∈ U. in the metric dH . Let g1 be the closed path in C which is the union of g and a segment connecting g(L) to g(0) in U .3) the time L geodesic flow restricted to U expands along the leaves of F uu and contracts along the leaves of F ss . and • on a fixed compact subset of Q1 M(S. Then there exists a unique closed geodesic g0 ⊂ C with the following properties: (I) g0 and g1 have lifts in T (S) which stay δ–close with respect to the Teich- müller metric. In view of Lemma 8. Given a quadratic differential (x. Then. we only need to consider the set of closed geodesics going through a fixed compact subset of C. q2 ) ≤ δ. • for any q1 . We can apply the contraction mapping principle to F ss to find (x0 . . g(L) ∈ U and (b) g spends more than half of its length in K. L] → C is a Teichmüller geodesic segment such that (a) g(0). • every nearly closed orbit approximates a close orbit (Lemma 9. there exist constants L0 > 0. q0 ) ∈ U 0 such that gL (q0 ) ∈ F uu q0 .2. We remark that in Lemma 9.4 there is in fact a neighborhood of (x. Consider the stable and unstable foliations for the geodesic flow. Our goal is to show that if U is small enough.1 if we remove the assumption that K consists of non-orbifold points then there are at most a uniformly bounded number of closed geodesics satisfying conditions (I–III). q) ∈ K and δ > 0. For L > L0 . the first return map on these foliations will define a contraction with respect to the Hodge distance function.1. this is the same as a closed geodesic going through U 0 .2. there are neighborhoods U ⊂ U 0 of (x. We have • by Theorem 3. σ) the geodesic flow is uniformly hyperbolic.  Proof of Theorem 1. A version of the closing lemma can be found in [H3]. if q1 ∈ F ss (q2 ) or q1 ∈ F uu (q2 ) then dH (q1 . COUNTING CLOSED GEODESICS IN STRATA 51 Lemma 9. Now we can consider the first return map of the map g−t on F uu (q0 ). Remark 9. q) such that the time L geodesic flow restricted to the neighborhood expands along the leaves of F uu and contracts along the leaves of F ss . (II) The length of g0 is within δ of L.1). L] → C satisfies properties (a) and (b) then in view of the hyper- bolicity statement (Lemma 8. Let K be a compact subset of C consisting of non- orbifold points.1. and open neighborhoods U ⊂ U 0 ⊂ C of (x. the geodesic flow on C is mixing. suppose that g : [0.3 and Lemma 8. Note that by the bound proved in Theorem 1. q) such that: • if g : [0.1 (Closing Lemma). q) (in a slightly bigger neighboorhood).5. q) with the following property. As a result. the contraction mapping principle (applied first to the map on F ss and then to the inverse of the map on F uu ) allows us to find a fixed point for the geodesic flow near (x. we find a fixed point for the first return map in U 0 .

Ergodic Theory Dynam. Mosc. 245–261. Duke Math. Systems. Yoccoz. K. Mod. Logarithmic asymptotics for the number of periodic orbits of the Teich- mueller flow on Veech’s space of zippered rectangles. Inventiones Math 153 (2003) 631–683. Masur. Funct. 2-3. Masur. 698–754. 71–105. Poénaru. and Dynamics I. Gardiner and H. J. 2. Hasselblat. and J. 141 (1978). 73–98. Hamenstädt. Paris Sér. Margulis. [FM] B. Bleiler and A. Steklov. [GM] F. Dynamics 4 (2010). (1992). no 1. of Math. [EM2] A. Mirzakhani. 292 (1981). Casson. no. no. Bernoulli Measures for the Teichmüller Flow. Exponential Mixing for the Teichmüller flow in the Space of Quadratic Differentials. [Bu] A. Zorich. [H1] U. Skinning Maps are Finite-to-one. 66 and 67 (1979). 1. Eskin and H. J. 144 (2008). 167 (1988). Coefficients of expansion of pseudo-Anosov homeomorphisms.)  References [AY] P. 75–78. IHES. no. 121–140. no 2. (2). 111–116. of Math. and V. Bowen’s construction for the Teichmüeller flow. 155 (2002). Issled. Asymptotic formulas on flat surfaces. Princeton Mathematical Series 49 (2012). Choi. (2) 147 (1998). Upper bounds and asymptotics in a quantitative version of the Oppenheim conjecture. Sem. Athreya and G. [Iv] N. Teichmüller theory and applications to geometry. Preprint. Mat. L. Ivanov. 3.2 following the work of Margulis [Mar]. [Du] D. (2011). Astérisque.6 in [KH]. Eskin and M. Rafi. [KZ] M. Arnoux and J. Bers. [Ath] J. 191. Preprint. Kontsevich and A. 18 (2008). Forni. Hubbard. F. Mirzakhani. Combinatorics of Train Tracks. 6. Journal of Modern Dynamics . Acta Math. [KH] A. no. B. R. A. Volume 9 of London Mathematical Society Student Texts. Geom. Fathi. Otdel. V. Automorphisms of surfaces after Nielsen and Thurston. 21 (2001). [AR] A. C. [AGY] A. Ann. Leningrad. Katok. Construction de difféomorphismes pseudo-Anosov. 9 (2009). 209–237. Exponential mixing for the Teichmüller flow. Bufetov. [BC] S. Cambridge University Press. 1055–1111. Zap. Bufetov. [FLP] A. Maria João Resende. 443–478. Cambridge. Dedicata 119 (2006). 1–103. Travaux de Thurston sur les surfaces. [CSR] Y. A. Lattice Point Asymptotics and Volume Growth on Teichmüller space. [ABEM] J. G. C. (1988). Margalit. Topol. Complex Variables Theory Appl. (2006). Cambridge University Press. all the ingredients are in place to drive Theorem 1. S. Nauchn. 393–418. A primer on mapping class groups. Eskin. [EM1] A. Avila. Quantitative recurrence and large deviations for Teichmüller geodesic flow. Lines of minima and Teichmülcer geodesics. Dynamics of the Teichmueller flow on compact invariant sets. no 6. 104 (2006). Forni. [H2] U. Deviation of ergodic averages for area-preserving flows on surfaces of higher genus. Extremal length geometry of Teichmüller space. S. Topology. Dumas. Eskin and M. [Be] L. Counting closed geodesics in moduli space. Avila. Publ. 1. 161 (2012). Hamenstädt. Geom. COUNTING CLOSED GEODESICS IN STRATA 52 Hence. Preprint. Penner. . Harer and R. Series. [EMM] A. Ann. Preprint. (1995). J. Anal. Farb and D. Mozes. 143–211. 93–141. (See also §20. J. An extremal problem for quasiconformal maps and a theorem by Thurston. Sci. Athreya. [H3] U. Duke Math. Laudenbach.-C. [AF] J. I Math. (LOMI). Annals of Mathematics Studies 125. Gouezel. Math. Matrix Editions. C. Athreya. Hamenstädt. Yoccoz. [HP] J. no. [Fo] G. Deviation of ergodic averages for rational polygonal billiards. 285-319. [Hu] J. Connected components of spaces of Abelian differentials with prescribed singularities. Inst. Acad. Introduction to the Modern Theory of Dynamical Systems. Math. 1. 16 (1991).

Anal 17 (2007). Math. of Math. (1970). [MS2] H. Papadopoulos and R. of Math. 2. Strebel. Smillie. II. [Mas] B. Penner. Interval exchange transformations and measured foliations. 68. 10 (1985). Ann. [St] K. 14 (2007). Y. 142. Amer. Bull. Math. [R4] K.. of Math. Ph. [PP2] A. Ann. Closed geodesics in the thin part of moduli space. Quadratic differentials. 1. Topology 19(1980). 310. Math. Minsky. Ann. Soc. 455–543. Lett. Math. 23– 41. A combinatorial model for the Teichmüller metric. Kerckhoff. 333–341. Penner. Geom. [Mi] Y. (2003).1 (1990). 134. Thick-thin decomposition of quadratic differentials. Veech. Comparison of hyperbolic and extremal lengths. [Pe] R.. 3 (Nov. pp. Math. Maskit. [Th] W. 3. Preprint. Thesis. 902–974. Thurston. A characterization of pseudo-Anosov foliations. no. Hierarchical structure. Rafi. 130 (1987). Papadopoulos and R. Masur and J. Ann. COUNTING CLOSED GEODESICS IN STRATA 53 [Ker] S. Acad.S.. 83 (1996). 441– 530. 936–959. [Mar] G. Margulis. Springer-Verlag (1984). Funct. Rafi. Extremal length estimates and product regions in Teichmüller space. 169–200. no. Springer. no. 1991). Duke Math. Res. Enumerating pseudo-Anosov foliations. no. [MM] H. . (2) 124 (1986). [MS1] H. no. Second Series.1. Pacific J. Lanneau. Ergebnisse der Math 5.. Minsky. (2) 115 (1982). (N. [La] E. 10 (2000). Sci.. 359–377. no. no. 289–307. Masur and J. 417–431. On the geometry and dynamics of diffeomorphisms of surfaces. 249–286. Pa- cific J. A construction of pseudo-Anosov homeomorphisms. [Ma] H. 2. 381–386. [R3] K. Penner. Funct. Hausdorff Dimension of Sets of Nonergodic Measured Folia- tions. 4. Masur.) 19 (1988). Smillie. On some aspects of the theory of Anosov flows. Rafi. Rafi. Trans.A. Masur. [R2] K. no. Number 1 (1993). Amer. Geom. Hyperbolicity in Teichmüller space. In preparation. (1988) 179–197. The Teichmüller geodesic flow. J. 2011. Quadratic differentials with prescribed singularities and pseudo-Anosov diffeomorphisms. Connected components of the strata of the moduli spaces of quadratic diferentials. preprint. 3. [Ve] W. No. Geometry of the complex of curves. Fenn. [PP1] A. Commentarii Mathematics Helevetici. [R1] K. 159–173. Soc. no. The asymptotic geometry of Teichmüller space. 2.D. Anal.