Geom. Funct. Anal. Vol.

29 (2019) 871–889
https://doi.org/10.1007/s00039-019-00502-7
Published online June 1, 2019
c 2019 Springer Nature Switzerland AG GAFA Geometric And Functional Analysis

GEODESIC CURRENTS AND COUNTING PROBLEMS

Kasra Rafi And Juan Souto

Abstract. For every positive, continuous and homogeneous function f on the space
of currents on a compact surface Σ, and for every compactly supported filling current
α, we compute as L → ∞, the number of mapping classes φ so that f (φ(α)) ≤ L. As
an application, when the surface in question is closed, we prove a lattice counting
theorem for Teichmüller space equipped with the Thurston metric.

1 Introduction
Let Σ be a compact orientable surface of negative Euler-characteristic and denote
by C = C(Σ) the associated space of currents, endowed it with the weak-*-topology
(see §2). The mapping class group

Map(Σ) = Homeo+ (Σ)/ Homeo0 (Σ) = Homeo+ (Σ)/ Homeo0 (Σ)

of the interior Σ = Σ\∂Σ of Σ acts on C by homeomorphism and the embedding

ML(Σ) → C(Σ) of the space of measured laminations into C is mapping class group
equivariant. In particular we can push forward the Thurston measure μThu on ML =
ML(Σ) to a mapping class group invariant measure on C.
For a continuous and homogeneous map f : C → R+ , define
  
m(f ) = μThu λ ∈ ML  f (λ) ≤ 1 .

We will be only interested in functions  f which are positive on the subset ML.
For such functions, the set {λ ∈ ML  f (λ) ≤ 1} is compact and thus m(f ) < ∞.
An important instance of such function is fα (  ) = i(α,  ) where i : C × C → R is
the intersection pairing and α is a filling current (ensuring fα is positive on ML).
Relaxing notation, we write for a filling current α
  
m(α) = μThu λ ∈ ML  i(α, λ) ≤ 1

This material is based upon work supported by the National Science Foundation under Grant No.
DMS-1440140 while the authors were in residence at the Mathematical Sciences Research Institute
in Berkeley, California, during the Fall 2016 semester. The first author was also partially supported
by NSERC Discovery Grant, RGPIN 435885.
872 K. RAFI, J. SOUTO GAFA

instead of m(fα ). Other important examples of positive, homogenous, and contin-

uous functions on the space of currents arise as extensions of lengths functions,
see [EPS16]. In the particular case that the length function comes from a point
X ∈ T (Σ) in Teichmüller space then we set
  
m(X) = μThu λ ∈ ML  X (λ) ≤ 1 .
The function X → m(X) is invariant under the action of the mapping class group
on Teichmüller space and thus descends to a function on moduli space M = M(Σ).
Let

mΣ = m(X) dvolwp ,
M(Σ)

be the integral of that function with respect to the Weil–Petersson volume form.
With all this notation in place we are ready to state our main theorem, a general
counting statement for the number of mapping class group orbits of a given current.
Main Theorem. Let Σ be a compact surface of genus g and n boundary compo-
nents, and let Σ = Σ\∂Σ be its interior. For a positive, continuous and homogeneous
map f : C(Σ) → R+ and a filling compactly supported current α ∈ Cc (Σ), we have
    
# φ ∈ Map(Σ)  f φ(α) ≤ L m(α) m(f )
lim 6g+2n−6
= .
L→∞ L mΣ
We can view the Main Theorem as an extension of an earlier result due to
Mirzakhani. In [Mir16] she proved that if X is a hyperbolic surface and γ a filling
closed curve, then
  
# φ ∈ Map(Σ)  X (φ(γ)) ≤ L m(X) nγ
lim 6g+2n−6
= (1.1)
L→∞ L mΣ
without providing the explicit value for nγ . In fact, Mirzakhani’s result is used as
key step in our proof and in section 3 we will show that nγ = m(γ).
Lattice counting in Teichmüller space. The Main Theorem can be applied
to other problems besides counting curves. As a perhaps unexpected application we
prove a lattice counting theorem in Teichmüller space. Recall first that in [ABEM12],
Athreya, Bufetov, Eskin and Mirzakhani proved that for any two X, Y ∈ T the
number of orbit points Map(Σ)·Y within Teichmüller distance R of X is asymptotic,
as R tends to ∞, to a constant multiple of edim T ·R . We establish the analogous result
when we endow Teichmüller space with the Thurston metric

dThu (X, Y ) = log inf Lip(h) .
h:X→Y

Here Lip(h) is the Lipschitz constant of h and the infimum is taken over all Lipschitz
maps h : X → Y in the right homotopy class. For X ∈ T and η ∈ C, define
i(μ, η)
DX (η) = max . (1.2)
μ∈ML X (μ)
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 873

Suppose now that the surface Σ is closed. Following Bonahon [Bon88] we can embed
Teichmüller space T as a subset of C in such a way that for every Y ∈ T and for
every curve γ we have
i(γ, Y ) = Y (γ).
We thus get from a result of Thurston [Thu86] that
dThu (X, Y ) = log DX (Y ).
Since the function DX : C → R+ is positive, homogeneous and continuous, we can
apply the Main Theorem for α = Y , f = DX . Letting L = eR , we obtain:
Theorem 1.1. Let Σ be a closed surface of genus g ≥ 2 an let X, Y ∈ T (Σ) be
two points in Teichmüller space. Then,
    
# φ ∈ Map(Σ)  dThu X, φ(Y ) ≤ R m(DX ) m(Y )
lim (6g−6)R
= ,
R→∞ e mΣ
where dThu is the Thurston metric on Teichmüller space and DX is as in (1.2). 
.
Erlandsson and the second author of this note have proved that Theorem 1.1
remains also true when the involved surface is not closed. This will appear in [ES].
Remark. Unlike the Teichmüller metric, the Thurston metric dThu is not symmet-
ric. This is reflected
 in the limiting constant of Theorem 1.1. On the other hand,
replacing dThu X, φ(Y ) ≤ R by dThu (φ(Y ), X) ≤ R will not make a difference
because
         
φ ∈ Map(Σ)  dThu φ(Y ), X ≤ R = φ ∈ Map(Σ)  dThu Y, φ−1 (X) ≤ R .
Convergence of measures. We now comment briefly on the proof of the main
Theorem. Consider the space M(C) of Radon measures on C, endowing it in turn with
the weak-*-topology induced by the algebra of continuous functions with compact
support Cc0 (C). In plain language, this means that, for νn , ν ∈ M(C),
 
νn → ν in M(C) ⇐⇒ f νn → f ν for every f ∈ Cc0 (C).
C C

Denoting by δα ∈ M(C) the Dirac measure centered at a compactly supported

current α ∈ Cc (Σ), we will derive the Main Theorem from the convergence of the
measures
1
ναL = 6g+2n−6 δ 1 f (α) ∈ M(C) (1.3)
L L
f ∈Map(Σ)
as L tends to ∞. We prove:
Theorem 1.2. With notation as in the Main Theorem we have
m(α)
lim ναL = · μThu .
L→∞ mΣ
874 K. RAFI, J. SOUTO GAFA

Remark. A version of this theorem, where α is assumed to be a curve appears in

[ES16], however, the constant there is not explicit.

The strategy to prove Theorem 1.2 is as follows. First we show that the fam-
ily (ναL )L is precompact (Lemma 4.1), meaning that every sequence of Ln → ∞
L
has a subsequence (Lni ) such that the measures να ni converge to some measure ν.
We prove next that any such limit ν is supported by the space ML of measured
laminations (Lemma 4.2), and that it gives measure 0 to the set of non-filling lami-
nations (Lemma 4.4). It then follows from Lindenstrauss–Mirzakhani’s classification
of locally finite mapping class group invariant measures on ML [LM08, Ham09] that
the limit ν is a multiple c · μThu of the Thurston measure (Lemma 4.5). All that
remains is to show that c = m(α)
mΣ . We deduce that this is the case, using a slightly
improved version of Equation (1.1): in Theorem 3.1 we show that the constant nγ
in Equation (1.1) equals m(γ). Mirzakhani’s proof of (1.1) relies on existence of the
limit
  
volwp X ∈ T  X (γ) ≤ L
lim .
L→∞ L6g+2n−6
We make use of results of Bonahon-Sözen [BS01] and Papadopoulos [Pap91] to
reprove the existence of this limit and further compute its value (Theorem 3.3).
Further comments. We add a few comments that the reader may find interesting.
1. Let Γ be a finite index subgroup of Map(Σ). The results of this paper also hold
for Γ up to dividing the constants by the index [Map(Σ) : Γ]. For instance, if α and
f are as in the Main Theorem, then
    
# φ ∈ Γ  f φ(α) ≤ L 1 m(α) m(f )
lim = · .
L→∞ L6g+2n−6 [Map(Σ) : Γ] mg
This is because everything we use, specifically the results in [Mir16], also holds for
Γ.
2. One should definitively point out that the results in this paper do not apply if
the surface Σ is not orientable. In fact, already Mirzakhani’s Equation (1.1) fails in
that case, [Gen, Mag].
3. The arguments in this paper show that not only does Mirzakhani’s (1.1) imply
Theorem 1.2 and thus the Main Theorem, but that also the implications can be
reversed. It would be interesting to see if the lattice counting theorem with respect
to the Teichmüller metric [ABEM12] is also equivalent to these results.

2 Currents
In this section, we recall a few facts on currents. See [AL, Bon86, Bon88] for a
thorough treatment.
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 875

Definitions and notation. Continuing with the same notation as in the intro-
duction, let Σ be a compact surface with negative Euler-characteristic. Let g be the
genus and n the number of boundary components of Σ.
We endow Σ with a fixed but otherwise arbitrary hyperbolic metric with respect
to which the boundary is geodesic and let P Trec Σ (resp. Trec 1 Σ) be the subset of

the projective tangent bundle P T Σ (resp. unit tangent bundle T 1 Σ) consisting of

lines (resp. vectors) tangent to bi-infinite geodesics. By construction, P Trec Σ has a
1–dimensional foliation whose leaves are the lifts of geodesics.
A geodesic current is a Radon transverse measure to the geodesic foliation of
P Trec Σ. Equivalently, a geodesic current is a π1 (Σ)-invariant Radon measure on the
space of unoriented geodesics in the universal cover of Σ. Also, geodesic currents
are in one-to-one correspondence with Radon measures on Trec 1 Σ which are invariant

under both the geodesic flow and the geodesic flip. Recall that a Borel measure is
Radon if it is locally finite and inner regular. We denote the space of all geodesic
currents, endowed with the weak-*-topology, by C(Σ).
It follows from the very definition of current that we can consider (unoriented)
periodic orbits of the geodesic flow, that is closed geodesics, as geodesic currents.
In fact, the set R+ S of weighted closed geodesics is dense in C(Σ). A measured
lamination is a lamination of Σ = Σ\∂Σ endowed with a transverse measure of
full support. As such, a measured lamination is also a current. Actually, the space
ML(Σ) of measured laminations is nothing but the closure in C(Σ) of the set of all
weighted simple curves other than the boundary. Thurston proved that ML(Σ) is,
with the induced topology, homeomorphic to R6g+2n−6 .

Mapping class group action. The space C(Σ) of currents is independent of the

chosen metric on Σ in the following sense: Any homeomorphism Σ → Σ between
hyperbolic surfaces with geodesic boundary induces a homeomorphism P Trec Σ →
 
P Trec Σ mapping one geodesic foliation to the other. Then the map C(Σ) → C(Σ )
between the spaces of currents induced by the foliation preserving homeomorphism

P Trec Σ → P Trec Σ is also a homeomorphism. Note in particular that the mapping
class group

Map(Σ) = Homeo+ (Σ)/ Homeo0 (Σ) = Homeo+ (Σ)/ Homeo0 (Σ)

of the interior Σ = Σ\∂Σ acts on C(Σ) by homeomorphisms.

The space ML(Σ) is not only homeomorphic to euclidean space, but has also
a compatible mapping class group invariant integral PL-manifold structure. Here,
integral means that the change of charts are given by linear transformations with
integral coefficients.

Compactly supported currents. We will be mostly interested in currents in

the set
 
C0 (Σ) = λ ∈ C(Σ) with λ(∂Σ) = 0 ,
876 K. RAFI, J. SOUTO GAFA

that is in those currents which do not assign any weight to the boundary components
of Σ. Also, for K ⊂ Σ compact let
 
CK (Σ) = λ ∈ C(Σ) with supp(λ) ⊂ GK (Σ)

be the set of currents supported by the set of GK (Σ) of geodesics in the universal
cover of Σ whose projection to the base is contained in K. Finally set

Cc (Σ) = CK (Σ)
K⊂Σ compact

to be the set of currents supported by some compact set. We refer to the element
of Cc (Σ) as compactly supported currents. We make a few comments on compactly
supported currents:
• While C(Σ) and CK (Σ) are projectively compact, neither C0 (Σ) not Cc (Σ) are
if ∂Σ = ∅. This is basically the reason why one needs to be more careful while
working with currents on open surfaces.
• For every K ⊂ Σ compact there is another compact set K  ⊂ Σ with
Map(Σ) · CK (Σ) ⊂ CK  (Σ).
In particular, both C0 (Σ) and Cc (Σ) are invariant under the mapping class
group.
• Let X be a hyperbolic metric on the interior Σ of Σ. The length function X on
the set of curves S extends continuously to a positive homogenous function on
the space Cc (Σ) of compactly supported currents. In fact, many other length
functions extend as well [EPS16].
• There is some compact set K ⊂ Σ with ML(Σ) ⊂ CK (Σ). In other words,
measured laminations are compactly supported currents.
Intersection pairing. In [Bon86] Bonahon introduced the map
i : C(Σ) × C(Σ) → R+ , i(λ, μ) = (λ ⊗ μ)(DP T Σ), (2.1)
where DP T Σ is the bundle over Σ whose fibre over p is the set
 
DP T Σp = (v, w) ∈ P Tp Σ × P Tp Σ with v = w

consists of pairs of distinct points in the projective space of the tangent space Tp Σ.
Bonahon proved that i(  ,  ) is continuous and homogenous, where homogenous means
that
i(t · λ, s · μ) = s · t · i(λ, μ)
for all λ, μ ∈ C(Σ) and t, s ∈ R+ . Moreover, it follows directly from the definition
that for any two closed primitive geodesics α, β in Σ the quantity i(α, β) is nothing
other than the minimal number of transversal intersections of curves homotopic to
α and β and in general position. The map i is called the intersection pairing.
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 877

While i(α,  ) = 0 whenever α is a boundary component of Σ, we have that the

intersection pairing is non-degenerate on the set C0 (Σ) of currents supported by the
interior of Σ. In fact, measured laminations are characterized to be those currents
in C0 (Σ) with vanishing self-intersection number:
 
ML(Σ) = λ ∈ C0 (Σ) with i(λ, λ) = 0 .

On the other extremum, a current λ ∈ C0 (Σ) is called filling if i(λ, μ) > 0 for every
non-zero current μ ∈ C0 (Σ). Equivalently, a current is filling if its support meets
transversally every bi-infinite geodesic other than the boundary components.
We list a few facts on i(  ,  ):

• If φ : Σ → Σ is a homeomorphism between hyperbolic surfaces, then the

induced homeomorphism φ : C(Σ) → C(Σ ) commutes with the scaling action
of R+ and with the intersection pairing. In particular, the intersection paring
is invariant under the mapping class group action Map(Σ)  C(Σ).
• If K ⊂ Σ is compact and if λ ∈ C0 (Σ) is a filling current, then the set
  
μ ∈ CK (Σ)  i(λ, μ) ≤ 1

is compact in CK (Σ).
• For any two filling currents μ, λ ∈ Cc (Σ) and for every compact set K ⊂ Σ,
there is some C with

C −1 · i(μ, η) ≤ i(λ, η) ≤ C · i(μ, η)

for every η ∈ CK (Σ).

• Suppose that μ ∈ Cc (Σ) is a filling current and that X ∈ T (Σ) is a complete
hyperbolic structure on the interior Σ of Σ. For every compact set K ⊂ Σ,
there is some C with

C −1 · i(μ, η) ≤ X (η) ≤ C · i(μ, η)

for every η ∈ CK (Σ).

Before moving on, suppose for a moment that Σ is closed, that is that it has
empty boundary. Then, for every point X ∈ T (Σ) in Teichmüller space, there is a
current, again denoted by X and called the Liouville current, with

i(X, γ) = X (γ)

for every closed curve γ. The Liouville current X is a filling current.

878 K. RAFI, J. SOUTO GAFA

A comment on the continuity of the i(  ,  ) In the literature, currents are

often either discussed on the setting of hyperbolic groups or for closed surfaces.
Considering currents on Σ is consistent with the former point of view. What is
specific about surfaces is the existence of the intersection pairing – continuity being
the only non-trivial result. We explain briefly how the continuity of the intersection
pairing follows from the result in the closed case. Let Σ̂ be a closed hyperbolic
surface for which there is an isometric embedding Σ → Σ̂. For instance, take Σ̂
to be the double of Σ. The inclusion of Σ into Σ̂ induces a foliation preserving
embedding P Trec Σ → P Trec Σ̂. This embedding induces an embedding C(Σ) → C(Σ̂).
By definition, the intersection pairing on C(Σ) is nothing other than the restriction
of the intersection pairing on C(Σ̂). Continuity of the latter then implies continuity
of the former.

3 Weil–Petersson Volume
In this section, we compute nγ in Equation (1.1) proving the following version of
the Mirzakhani’s theorem [Mir16, Theorem 1.1].
Theorem 3.1. Let Σ be a compact surface of genus g with n boundary components
and let X be a complete finite volume hyperbolic structure on its interior Σ = Σ\∂Σ.
Then   
# φ ∈ Map(Σ)  X (φ(γ)) ≤ L m(X) m(γ)
lim 6g+2n−6
=
L→∞ L mΣ
for every filling closed curve γ ⊂ Σ.
We need some preparation before coming to the proof. As we mentioned earlier,
ML is naturally a PL-manifold. Moreover, it is endowed with a mapping class
group invariant symplectic structure, namely Thurston’s intersection pairing [PH92].
Actually, this symplectic structure is compatible with the Weil–Petersson form on T .
More precisely, let μ be a geodesic lamination where the complementary components
are ideal triangles and let ML(μ) be the open and dense subset of ML consisting
of measured laminations that are transverse to μ. Note that μ itself may not be a
measured lamination. In fact, in case Σ has boundaries, we take (for simplicity) μ to
be a union of finitely many bi-infinite geodesics. Thurston has showed [Thu86] that
there is a global parametrization
Φμ : T → ML(μ).
sending a hyperbolic structure X to a measured lamination associated to the mea-
sured foliation F transverse to μ such that the F –measure of a sub-arc ω of μ is
the hyperbolic length of ω in X. The map Φμ is relevant in the present setting
becauseit was shown by Bonahon and Sözen proved in the closed case [BS01] and
by Papadopoulos and Penner in the case where μ is finite [PP93] that it is a sym-
plectomorphism when the source is endowed with the Weil–Petersson form and the
target with Thurston’s intersection pairing.
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 879

Now, the symplectic structure on ML induces a mapping class group invariant

measure in the Lebesgue class. This measure is the so-called Thurston measure μThu .
It is an infinite but locally finite measure, positive on non-empty open sets. Further
we have

μThu (L · U ) = L6g−6 · μThu (U )

for all U ⊂ ML(Σ) and L > 0. This implies that μThu (A) = 0 if A ∩ L · A = ∅ for
all L > 0. In particular,
  
μThu λ ∈ ML(Σ)  f (λ) = L = 0

for any positive continuous homogenous function f on the space of currents. Finally,
note that ML(μ) has full measure.

Remark. It is due to Masur [Mas85] that, up to scaling, the Thurston measure is

the only Map(Σ) invariant measure on ML in the Lebesgue class. However, several
natural normalization are possible and it is not known if they agree. For example,
one could normalize the measure to be equal to the scaling limit
1
lim δ1γ
L→∞ L6g−6 L
γ∈MLZ

of the counting measure on integral multi-curves. For a discussion of how the measure
defined by the symplectic structure (which is what we use in this paper) is related to
the measure defined by the scaling limit (used by Mirzakhani in [Mir16]) see [MT19]

The main idea in [Mir16] is to relate the counting problem in (1.1) to the Weil–
Petersson volume of certain sets in T . More precisely, for a filling curve γ and L > 0,
Mirzakhani considered the sets


BT (γ, L) = X ∈ T  X (γ) ≤ L

and showed [Mir16, Theorem 8.1] that

 
volwp BT (γ, L)
lim (3.1)
L→∞ L6g−6
exists. We reprove this theorem, further giving a precise value for the limit. First,
we recall the follow lemma from [Pap91, Lemma 4.9].

Lemma 3.2. (Papadopoulos) Let Xn be a sequence of points in T converging to

a lamination λ in the Thurston boundary. Then, for every curve α, there exists a
constant C such that for every n, we have
   
i Φμ (Xn ), α ≤ α (Xn ) ≤ i Φμ (Xn ), α + C.
880 K. RAFI, J. SOUTO GAFA

Theorem 3.3. We have

 
volwp BT (γ, L)
lim = m(γ)
L→∞ L6g−6
for any filling curve γ.

Proof. Recall that the map Φμ considered earlier is a symplectomorphism and thus
volume preserving. Thus, to compute the Weil–Petersson  of BT (γ, L), we
volume
can instead find out the Thurston volume of BL = Φμ BT (γ, L) :

volwp (BT (γ, L)) = μThu (BL ).

Note that it follows from the convexity of lengths with respect to the shearing
coordinates [BBFS13, The14] that BL is a closed convex set in ML(μ).
Since Thurston’s volume form is compatible with the linear structure in ML, we
have

6g−6 1
μThu (BL ) = L μThu BL
L
where
 
1 1 
BL = λ  λ ∈ BL .
L L
Also, it follows immediately from Lemma 3.2 that, for sequences Xn ∈ T and
Ln > 0 and a measure lamination λ ∈ ML, if L1n Xn → λ ∈ ML as a sequence of
geodesic currents in C(S), then
1
Φμ (Xn ) → λ.
Ln
1
This implies that the family of sets L BL
converge point-wise to the set


BML (γ, 1) = λ ∈ ML(μ)  i(γ, λ) ≤ 1 .

That is, we have a family of closed convex sets that converge point-wise to a closed
set. Then their volume converges as well. Altogether we get

1 1 1
6g−6
volwp (BT (γ, L)) = 6g−6 μThu (BL ) = μThu BL −→ BML (γ, 1).
L L L

And m(γ) is nothing but the Thurston volume of BML (γ, 1) because ML(μ) has
full measure. 


Mirzakhani derived Equation (1.1) from the existence of the limit (3.1). To obtain
Theorem 3.1 we need to show:
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 881

Theorem 3.4. For any filling curve γ,

     m(X)
# φ ∈ Map(Σ)  X (φ(γ)) ≤ L ∼ volwp BT (γ, L) · , (3.2)
mΣ
where ∼ means that the ratio tends to 1 when L → ∞.
Proof. This is essentially proven as part of proof of Theorem 1.1 in [Mir16] (see
[Mir16, Section 9.4]). Let P = {α1 , . . . , α3g−3 } be a pants decomposition of Σ
and consider the the associated Fenchel–Nielsen coordinates for Teichmüller space,
namely
 
X → 1 (X), τ1 (X), . . . , 3g−3 (X), τ3g−3 (X) .

where i (X) is the length of αi at X and τi (X) is the amount of twisting around
the curve αi . For m = (m1 , . . . , m3g−3 ) ∈ Z3g−3 define CPm to be the cone in this
coordinate consisting of all points X ∈ T where
mi · i (X) ≤ τi (X) ≤ (mi + 1) · i (X).
Now, [Mir16, Equation 9.4] states that
     m(X)
# φ ∈ Map(Σ)  φ(X) ∈ CPm ∩ BT (γ, L) ∼ volwp CPm ∩ BT (γ, L) · . (3.3)
mΣ
That is, the theorem holds for every cone. However, we need infinitely many such
cones to cover BT (γ, L) and we need to be a bit careful.
As one applies larger and larger powers of the Dehn twist around the curve αi ,
the length of γ (which has to intersect αi because γ is filling) eventually becomes
larger than τi (X) · i (X). Therefore, there is a constant E0 = E0 (γ, X) depending
on γ and X so that, for L large enough,
L
φ(X) ∈ CPm ∩ BT (γ, L) =⇒ αi (φ(X)) ≤ E0 · .
mi
Using the fact that wp–volume form can be written as d1 · · · d3g−3 dτ1 · · · dτ3g−3
and setting E2 = E06g−6 , we also have
  L6g−6
volwp CPm ∩ BT (γ, L) ≤ E2 . (3.4)
m21 m22 · · · m23g−3
Recall [Mir16, Lemma 5.1] which states that, for some E1 depending on E0 ,
 L L6g−6
# φ(P )  φ(P ) (X) ≤ L, ∀i φ(αi ) (X) ≤ ≤ E1 .
mi m21 m22 · · · m23g−3
This can be restated as
   L6g−6
# φ ∈ Map(Σ)  φ(X) ∈ CPm ∩ BT (γ, L) ≤ E1 , (3.5)
m21 m22 · · · m23g−3
882 K. RAFI, J. SOUTO GAFA

because, for every φ in the first set, a composition φ−1 with the correct number
of Dehn twists around the curves αi is in the second set and this gives a bijection
between the two sets. But
1 1 1
2 × 2 × ··· × 2 ≤ ∞.
m1 m2 m3g−3
m∈N3g−3

Therefore, for every , there exists a finite collection of cones CPm so that the
proportion of the contribution to both sides of Equation (3.2) from all the other
cones is less than , independent of the value of L. Hence, Equation (3.3) implies
Equation (3.2). 


Armed with Theorem 3.1, we can obtain a special case of the Main Theorem
where the function f is given by intersection number and where, more crucially, the
current is assumed to be a filling curve. This simpler statement will be used later in
the proof of the Main Theorem.

Corollary 3.5. For every filling curve γ in Σ and for every filling current α ∈ C
we have
  
# φ ∈ Map(Σ)  i(α, φ(γ)) ≤ L m(γ)m(α)
lim = .
L→∞ L6g−6 mg

Proof. Recall, from [ES16, Corollary 4.4], that

  
# φ ∈ Map(Σ)  i(α, φ(γ0 )) ≤ L m(α)
lim    =

L→∞ # φ ∈ Map(Σ) i(X, φ(γ)) ≤ L m(X)

for any X ∈ T . The claim follows from this and Theorem 3.1. 


4 Limit of Measures
In this section we prove Theorem 1.2. Recall that M(C) is the space of Radon mea-
sures on C = C(Σ), endowed with the weak-*-topology. Also, fix a filling compactly
supported current α ∈ Cc (Σ) and recall the definition of the measures
1
ναL = δ 1 φ(α) ∈ M(C)
L6g+2n−6 L
φ∈Map(Σ)

from the introduction. The first step in the proof of Theorem 1.2 is to show that
the family of measures ναL is precompact, meaning that any sequence Ln → ∞ has
L
a subsequence (Lni ) such that (να ni )i converges.

Lemma 4.1. The family (ναL )L is precompact in M(C) and every accumulation point
is locally finite and positive.
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 883

Proof. First of all, we fix some finite area hyperbolic structure X on Σ. We also fix

a filling curve γ0 and some C ≥ 1 with
C −1 · i(λ, γ0 ) ≤ i(λ, α) ≤ C · i(λ, γ0 ) (4.1)
for every current λ ∈ Cc (Σ).
Note now that there is some compact set K ⊂ Σ = Σ\∂Σ with φ(α) ∈ K for
all φ ∈ Map(Σ). It follows that L
 να is supported on CK (Σ) for all K. This is a

key fact because {λ ∈ CK (Σ) X (λ) ≤ T } is compact for all T ≥ 0 and in the
weak-*-topology on a compact space, the space of probability measures is compact.
Therefore, precompactness of the family (ναL )L follows once we prove that
  
lim sup ναL {λ ∈ CK (Σ)  X (λ) ≤ T } < ∞
L→∞

for every T ≥ 0. Let us compute:

   
   1  1
ναL {λ ∈ CK (Σ)  X (λ) ≤ T } = # φ ∈ Map(Σ)   X φ(α) ≤ T
L6g+2n−6 L
1   
= # φ ∈ Map(Σ)  i(X, φ(α)) ≤ T · L
L6g+2n−6
1   
= # φ ∈ Map(Σ)  i(φ−1 (X), α) ≤ T · L
L 6g+2n−6

1   
≤ # φ ∈ Map(Σ)  i(φ−1 (X), γ0 ) ≤ C · T · L
L6g+2n−6
1   
= # φ ∈ Map(Σ)  i(X, φ(γ0 )) ≤ C · T · L
L6g+2n−6
1   
= # φ ∈ Map(Σ)  X (φ(γ0 )) ≤ C · T · L .
L 6g+2n−6

Now, from Theorem 3.1 we get that, when L grows, the last quantity approaches
the quantity
m(γ0 )m(X)
(C · T )6g+2n−6 .
mΣ
  
In particular, the function L → ναL λ ∈ CK (Σ)  X (λ) ≤ T is bounded, as we
needed to show.
Note that we have also already proved that every accumulation point of (ναL )L
is locally finite. To prove that any such accumulation point is positive, it suffices to
prove conversely that
  
lim inf ναL λ ∈ CK (Σ)  X (λ) ≤ 1 > 0.
L→∞

A completely analogous calculation as above, using this time the second inequality
of (4.1) implies that
  
   # φ ∈ Map(Σ)  X (φ(γ0 )) ≤ K −1 · L
L 
να ( λ ∈ C X (λ) ≤ 1 ) ≥ .
L6g+2n−6
The claim follows, once again, from Theorem 3.1. 

884 K. RAFI, J. SOUTO GAFA

Lemma 4.1 asserts that (ναL )L has accumulation points. Over the next few lemmas
we will study such points.

Lemma 4.2. Suppose that Ln → ∞ is a sequence such that the limit

ν = lim ναLn
n→∞

exists in M(C). Then ν is supported by the subspace ML ⊂ C of measured lamina-

tions.

Proof. Fix again an auxiliary hyperbolic structure X ∈ T . In order to prove that ν

is supported by ML it suffices to prove that

i(λ, λ) ν(λ) = 0 (4.2)
{λ∈C | X (λ)≤L}

for all L. Computing the corresponding quantity for ναL we get

 
L 1 φ(α) φ(α)
i(λ, λ) να (λ) = 6g+2n−6 i ,
{λ∈C | X (λ)≤L} L L L
φ∈Map(Σ), X (φ(α))≤L
1 i(α, α)
=
L6g+2n−6 L2
φ∈Map(Σ), X (φ(α))≤L
  
# φ ∈ Map(Σ)  X (φ(α)) ≤ L i(α, α)
= 6g+2n−6
·
L L2
   i(α, α)
= ναL λ ∈ C  X (λ) ≤ T ·
L2
where the second equality holds because of the homogeneity and invariance of the
intersection pairing. The third equality holds because all the summands are identical
and the fourth is just the definition of ναL . Recall that, in the proof of Lemma 4.1,
we have shown

lim sup ναL ({λ ∈ C  X (λ) ≤ T }) < ∞.
L→∞

Plugging this in the previous computation we deduce that


lim i(λ, λ)ναL (λ) = 0,
L→∞ {λ∈C | X (λ)≤L}

from where (4.2) follows. 



We know at this point that every accumulation point ν of the family (ναL )L is
supported by the space a measured laminations. We will prove below that any such
ν is a multiple of the Thurston measure. It is known [ES16, Proposition 4.1]—and
this is going to be important in the next Lemma—that this is the case if the filling
current α is a filling curve:
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 885

Proposition 4.3. Let η be a filling multicurve and suppose that Ln → ∞ is a

sequence such that the limit ν = limn→∞ νηLn exists in M(C). Then ν = c · μThu for
some c > 0. 

For now, we prove that, for a general filling current, any accumulation point of
(ναL ) when L → ∞ gives measure 0 to the set of laminations which fail to intersect
some curve.
Lemma 4.4. Suppose that Ln → ∞ is a sequence such that the limit ν =
limn→∞ ναLn exists. Then we have
  
ν λ ∈ ML  i(λ, γ) = 0 = 0
for every simple curve γ ⊂ Σ.
Proof. Start by fixing an auxiliary hyperbolic structure X ∈ T (Σ) and some C ≥ 1
with
C −1 · i(X, λ) ≤ i(α, λ) ≤ C · i(X, λ)
for every current λ ∈ C. The claim of the lemma follows when we show that for all
T > 0 we have
  
lim lim sup ναL λ ∈ C  i(λ, γ) ≤ T and X (λ) ≤ T = 0. (4.3)
→0 L→∞

Before launching the proof of (4.3), choose a filling curve η and note that, up to
increasing K just this time only, we can further assume that
K −1 · i(η, λ) ≤ i(α, λ) ≤ K · i(η, λ)
holds again for every current λ ∈ C.
Computing basically as above we get the following:
  
ναL λ ∈ C  i(λ, γ) ≤ T and X (λ) ≤ T
1   
= 6g+2n−6 # φ ∈ Map(Σ)  i(φ(α), γ) ≤ T L, X (φ(α)) ≤ T L
L
1   
≤ 6g+2n−6 # φ ∈ Map(Σ)  i(φ(η), γ) ≤ T LK, X (φ(η)) ≤ T LK
L 
= νηL ({λ ∈ C  i(λ, γ) ≤ T K and X (λ) ≤ T K}).
From Proposition 4.3 we get then that any limit of νηL is a multiple of the Thurston
measure. In other words, there is c with
  
lim ναL λ ∈ C  i(λ, γ) ≤ T and X (λ) ≤ T ≤
  
≤ c · μThu λ ∈ C  i(λ, γ) ≤ T K and X (λ) ≤ T K
for all L. Since μThu gives vanishing measure to the set of those laminations which
do not intersect γ we get that this last quantity tends to 0 as  → 0. We have proved
(4.3) and thus Lemma 4.4. 

886 K. RAFI, J. SOUTO GAFA

We recall that Lindenstrauss and Mirzakhani [LM08] have classified all mapping
class group invariant measures on ML (see also [Ham09]). The following charac-
terisation of the Thurston measure is just a reformulation of Theorem 7.1 in their
paper:

Theorem (Lindenstrauss–Mirzakhani). Let ν be a locally finite Map(Σ)-invariant

measure on ML(Σ) with
  
ν λ ∈ ML(Σ)  i(γ, λ) = 0 = 0

for every simple closed curve γ ⊂ Σ. Then ν is a multiple of the Thurston measure
μThu .

We are now ready to prove that Proposition 4.3 also holds for arbitrary filling
currents α. In other words, we prove that any accumulation point of (ναL ) is a multiple
of the Thurston measure:

Lemma 4.5. Suppose that Ln → ∞ is a sequence such that the limit ν =

limn→∞ ναLn exists. Then ν = c · μThu for some c > 0.

Proof. Note that the limiting measure ν is positive and locally finite by Lemma 4.1.
Moreover, ν is supported by ML by Lemma 4.2. Also, since all the measures ναLn are
Map(Σ)-invariant we get that ν is Map(Σ)-invariant as well. Finally, by Lemma 4.4
we have
  
ν λ ∈ ML  i(λ, γ) = 0 = 0

for every simple curve γ ⊂ Σ. The claim follows from the Lindenstrauss–Mirzakhani
theorem. 


We are now ready to prove Theorem 1.2:

Proof of Theorem 1.2. By Lemma 4.1 we know that the any sequence Ln → ∞ has
a subsequence (Lni ) such that the limit

ν = lim ναLn
n→∞

exists. Then, by Lemma 4.5 we know that ν = c · μThu for some constant c = c(Lni )
depending on the sequence (Lni ). To prove the theorem, it suffices to show that

m(α)
c= , (4.4)
mg
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 887

which implies in particular that c is independent of the sequence Lni . To that end
fix a curve γ and compute:
   L   
ν λ ∈ C  i(λ, γ) ≤ 1 = lim να ni λ ∈ C  i(λ, γ) ≤ 1
i→∞
1   
= lim 6g+2n−6 # φ ∈ Map(Σ)  i(φ(α), γ) ≤ Lni
i→∞ Ln
i

1   
= lim # φ ∈ Map(Σ)  i(α, φ(γ)) ≤ Lni
6g+2n−6
i→∞ Ln
i

m(γ)m(α)
= ,
mg
where the last line holds by Corollary 3.5. Since ν = c · μThu we have
  
ν λ ∈ C  i(λ, γ) ≤ 1 = c · m(γ),
meaning that
m(γ)m(α)
c · m(γ) = .
mg
Equation (4.4) follows and we have proved Theorem 1.2. 


5 Proof of the Main Theorem

We restate the Main Theorem from the introduction.
Main Theorem. Let Σ be a compact surface of genus g and n boundary compo-
nents, and let Σ = Σ\∂Σ be its interior. For a positive, continuous and homogeneous
map f : C(Σ) → R+ and a filling compactly supported current α ∈ Cc (Σ), we have
    
# φ ∈ Map(Σ)  f φ(α) ≤ L m(α) m(f )
lim = .
L→∞ L6g+2n−6 mΣ
Proof. For L > 0, consider again the measures
1
ναL = 6g+2n−6 δ 1 φ(α) ,
L L
φ∈Map(Σ)

recall that they are supported by CK (Σ) for some K ⊂ Σ compact, and note that
for all L one has
⎛ ⎞
     
# φ ∈ Map(Σ)  f (φ(α)) ≤ L = ⎝ δφ(α) ⎠ η ∈ CK (Σ)  f (η) ≤ L
φ∈Map(Σ)
⎛ ⎞
  
=⎝ δ 1 φ(α) ⎠ η ∈ CK (Σ)  f (η) ≤ 1
L
φ∈Map(Σ)
  
= L6g+2n−6 · ναL η ∈ CK (Σ)  f (η) ≤ 1 .
888 K. RAFI, J. SOUTO GAFA

By Theorem 1.2, the limit of ναL as L goes to infinity is m(α)

mg μThu . We have

#{φ ∈ Map(Σ)  f (φ(α)) ≤ L} m(α)   
lim 6g+2n−6
= μThu η ∈ CK (Σ)  f (η) ≤ 1
L→∞ L mg
m(α)m(f )
= .
mg
This finishes the proof. 


Acknowledegements
The work of Maryam Mirzakhani provides the foundation for the results in this
paper and in general she has been an inspiration for us. We dedicate this paper,
which hopefully she would have found amusing, to her memory. We also thank the
referee for helpful comments.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional

claims in published maps and institutional affiliations.

References
[AL] J. Aramayona and C. Leininger, Hyperbolic structures on surfaces and
geodesic currents, to appear in Algorithms and geometric topics around auto-
morphisms of free groups. Advanced Courses CRM-Barcelona, Birkhäuser.
[ABEM12] J. Athreya, A. Bufetov, A. Eskin and M. Mirzakhani, Lattice Point
Asymptotics and Volume Growth on Teichmuller space. Duke Mathematical
Journal 161 (2012), 1055.
[BBFS13] M. Bestvina, K. Bromberg, K. Fujiwara and J. Souto, Shearing coordi-
nates and convexity of length functions on Teichmüller space. American Journal
of Mathematics 135 (2013).
[Bon86] F. Bonahon, Bouts des variétés hyperboliques de dimension 3. Annals of Math-
ematics 124 (1986).
[Bon88] F. Bonahon, The geometry of Teichmüller space via geodesic currents. Inven-
tiones Mathematicae 92 (1988).
[BS01] F. Bonahon and Y. Sözen, The Weil–Petersson and Thurston Symplectic
forms. Duke Mathematical Journal (3) 108 (2001), 581–597.
[EPS16] V. Erlandsson, H. Parlier and J. Souto, Counting curves and the stable
length of currents, arXiv:1612.05980, (2016).
[ES16] V. Erlandsson and J. Souto, Counting curves in hyperbolic surfaces. GAFA
26 (2016).
[ES] V. Erlandsson and J. Souto, Counting curves in orbifolds, in preparation.
[Gen] M. Gendulphe, What’s wrong with the growth of simple closed geodesics on
nonorientable hyperbolic surfaces, arXiv:1706.08798.
[Ham09] U. Hamenstädt. Invariant Radon measures on measured lamination space.
Inventiones Mathematicae (2)176 (2009), 223–273.
GAFA GEODESIC CURRENTS AND COUNTING PROBLEMS 889

[LM08] E. Lindenstrauss and M. Mirzakhani, Ergodic theory of the space of mea-

sured laminations. IMRN 4, (2008).
[Mag] M. Magee, Counting one sided simple closed geodesics on Fuchsian thrice punc-
tured projective planes, arXiv:1705.09377.
[Mas85] H. Masur, Ergodic actions of the mapping class group. Proceedings of AMS
94, (1985).
[Mir16] M. Mirzakhani, Counting mapping class group orbits on hyperbolic surfaces,
arXiv:1601.03342, (2016).
[MT19] L. Monin and V. Telpukhovskiy, On normalizations of Thurston measure
on the space of measured laminations, arXiv:1902.04533 (2019).
[PH92] R. Penner and J. Harer, Combinatorics of Train Tracks, Princeton University
Press, Princeton (1992).
[Pap91] A. Papadopoulos, On Thurston’s boundary of Teichmüller space and the
extension of earthquakes. Topology and Its Applications (3) 41 (1991), 147–
177.
[PP93] A. Papadopoulos and R. C. Penner, The Weil–Petersson symplectic struc-
ture at Thurston’s boundary. Transactions of the American Mathematical Soci-
ety 335 (1993), 891–904.
[The14] G. Théret, Convexity of length functions and Thurston’s shear coordinates,
arXiv:1408.5771 (2014).
[Thu86] W. Thurston, Minimal stretch maps between hyperbolic surfaces,
arXiv:math.GT/9801039, (1986).

Kasra Rafi, Department of Mathematics, University of Toronto, 40 St. George Street,

Toronto, ON M5S 2E4, Canada. [email protected]
[email protected]
Juan Souto, IRMAR - Université de Rennes 1, Beaulieu - Bâtiment 22-23, 263,
Avenue du Général Leclerc, 35042 Rennes, France. [email protected]

Received: March 30, 2018

Revised: November 12, 2018
Accepted: March 18, 2019