Abstract. Let Σ be a hyperbolic surface. We study the set of curves on Σ of a given type, i.e. in the mapping class group orbit of some fixed but otherwise arbitrary γ0. For example, in the particular case that Σ is a once-punctured torus, we prove that the cardinality of the set of curves of type γ0 and of at most length L is asymptotic to L 2 times a constant.
1.Throughout this paper we let Σ be a complete hyperbolic surface of finite area, with genus g and r punctures, and distinct from a thrice punctured sphere. By an immersed multicurve, or simply multicurve, in Σ we will mean an immersed compact 1-dimensional submanifold of Σ each of whose components represents (the conjugacy class of) a primitive non-peripheral element in π 1 (Σ). Two multicurves γ, γ are of the same type if they belong to the same mapping class orbit, meaning that there is a diffeomorphism φ of Σ such that γ and φ(γ ) are isotopic as immersed submanifolds. In general, isotopic geodesics are considered to be equivalent. For instance, every multicurve γ is isotopic to a geodesic multicurve and the length Σ (γ) is the length of the latter.In this paper we study the set S γ 0 = Map(Σ) · γ 0 of (isotopy classes of) multicurves of some given type γ 0 . More precisely, we are interested in the behavior, when L tends to infinity, of the number |{γ ∈ S γ 0 | Σ (γ) ≤ L}| of multicurves in Σ of type γ 0 and of at most length L. Since this number grows coarsely like a polynomial of degree 6g − 6 + 2r (see [18] for the case that γ 0 is simple and [20,21] or Corollary 3.6 below for the general case), the perhaps most grappling question is whether the limitexists. Our main result is that it does if Σ is a once-punctured torus: Theorem 1.1. Let Σ be a complete hyperbolic surface of finite volume homeomorphic to a once punctured torus and let γ 0 ⊂ Σ be a multicurve. The limit (1.1) exists and moreover we have where µ Thu is the Thurston measure on the space of measured laminations ML(Σ) and C γ 0 > 0 depends only on γ 0 .In the case of simple multicurves Theorem 1.1 is due to . Also, for simple multicurves, Mirzakhani [13] proved that the limit (1.1) exists for all g and r. Building on the work of Mirzakhani, Rivin [19] established the existence of the limit (1.1) for multicurves with a single self-intersection.Remark. Recently, and independently of our work, Mirzakhani [14] has established the existence of (1.1) in complete generality. Her argument and ours are different in nature and in some sense complementary. See the remarks following the statement of Corollary 4.4 in this introduction for more on the relation between Mirzakhani's result and ours.Still in the setting of simple multicurves, the case of the torus is much more treatable than the general one because any two simple multicurves in the torus are of the same type as long as they have the same number of components. This means that, in the case of the torus, counting simple multicurves of some fixed type basically reduces to counting all simple multicurves, a much simpler problem. In fact,...