You are on page 1of 49

Simple geodesics and Weil-Petersson volumes of

moduli spaces of bordered Riemann surfaces
Maryam Mirzakhani

July 12, 2005

Contents
1 Introduction 1

2 Background material 8

3 Geometry of pairs of pants 13

4 Generalized McShane identity for bordered surfaces 18

5 Statement of the recursive formula for volumes 27

6 Polynomial behavior of the Weil-Petersson volume 30

7 Leading coefficients of volume polynomials 34

8 Integration over the moduli space 36

9 Volumes of moduli spaces of bordered Riemann surfaces 43

1 Introduction
In this paper we investigate the Weil-Petersson volume of the moduli space of
curves with marked points. We develop a method for integrating geometric
functions over these moduli spaces, and obtain an effective recursive for-
mula for the volume Vg,n (L1 , . . . , Ln ) of the moduli space Mg,n (L1 , . . . , Ln )
of hyperbolic Riemann surfaces of genus g with n geodesic boundary compo-
nents. We show that Vg,n (L) is a polynomial whose coefficients are rational
multiples of powers of π. The constant term of the polynomial Vg,n (L) is

1

the Weil-Petersson volume of the traditional moduli space of closed surfaces
of genus g with n marked points.
In forthcoming papers, we will use these results to investigate problems
related to the distribution of the lengths of simple closed geodesics on hyper-
bolic surfaces, volume of −thin part of the moduli space and intersection
theory on moduli spaces of curves.
Volume of the moduli space. When studying volumes of moduli spaces of
hyperbolic Riemann surfaces with cusps, it proves fruitful to consider more
generally bordered hyperbolic Riemann surfaces with geodesic boundary
components. Given L = (L1 , . . . , Ln ) ∈ (R≥0 )n , the mapping class group
Modg,n acts on the Teichmüller space Tg,n (L) of hyperbolic structures with
geodesic boundary components of length L1 , . . . , Ln . We study the Weil-
Petersson volume of the quotient space

Mg,n (L) = Tg,n (L)/ Modg,n .

Our main result, obtained in §6, is:

Theorem 1.1. The volume Vg,n (L1 , . . . , Ln ) = Volwp (Mg,n (L)) is a poly-
nomial in L1 , . . . , Ln ; namely we have:

Vg,n (L) = Cα · L 2α ,
α
|α|≤3g−3+n

where Cα > 0 lies in π 6g−6+2n−|2 α| · Q.

Here the exponent α = (α1 , . . . , αn ) ranges over elements in (Z≥0 )n ,

n
Lα = Lα1 1 · · · Lαnn , and |α| = αi .
i=1
Moreover, in §5 we give an explicit recursive formula for calculating these
volumes. For example, we have:

V1,1 (L) = L2 /24 + π 2 /6.

For more examples see Table 1.
In particular, the Weil-Petersson volume of the moduli space of curves
of genus g with n marked point, the constant term of Vg,n (L), is a rational
multiple of π 6g−6+2n . This result was previously obtained by S. Wolpert
[Wol2]. A formula for Vol0,n (0), the Weil-Petersson volume of M0,n , was
obtained in [Zo].
Remark. Note that there is a difference in the normalization of the volume
form; in [Zo] the Weil-Petersson Kähler form is 1/2 the imaginary part of

2

Table 1. Volumes of moduli spaces of curves

g n Vg,n (L)

0 3 1

1 2
1 1 24 (L + 4π 2 )

1 2
0 4 2 (4π + L21 + L22 + L23 + L24 )

1 2
1 2 192 (4π + L21 + L22 )(12π 2 + L21 + L22 )

1
 
5  
5 
0 5 8 80π 4 + L4i + 4 L2i L2j + 24π 2 L2i
i=1 1≤i<j≤5 i=1

1

2 1 2211840 4π 2 + L21 ) (12π 2 + L21 ) (6960π 4 + 384 π 2 L21 + 5 L41 )

the Weil-Petersson pairing, while here we work with the imaginary part of
the pairing. So our answers are different by a power of 2.

We approach the calculation of these volumes by studying the lengths of
simple closed geodesics on X ∈ Mg,n . Our main tool is a generalization of
McShane’s identity [M], which gives us a way to calculate the volume of the
moduli space Mg,n = Tg,n / Modg,n without having to find a fundamental
domain for the action of the mapping class group on Teichmüller space.
McShane identity. Our point of departure for calculating these volume
polynomials is the following result [M]:

Theorem 1.2 (McShane). Let X be a hyperbolic once-punctured torus.
Then we have  1
(1 + eγ (X) )−1 = , (1.1)
γ
2

3

where the sum is over all simple closed geodesics γ on X.
Calculation of Vol(M1,1 ). We briefly explain the relation between Mc-
Shane’s identity and Weil-Petersson volumes by treating the case g = n = 1.
Consider the space of pairs:

M∗1,1 = {(X, γ) | X ∈ M1,1 , γ a simple closed geodesic on X},

and let
π : M∗1,1 → M1,1
be the projection map, defined by π(X, γ) = X. Also, define  : M∗1,1 → R
by
(X, γ) = γ (X).
Then we can rewrite (1.1) as
 1
f ((Y )) = , (1.2)
2
π(Y )=X

where f (x) = (1 + ex )−1 .
For any simple closed curve α on a hyperbolic once punctured torus, we
have M∗1,1 = T1,1 / Stab(α). Now we use the Fenchel-Nielsen coordinates for
T1,1 about α; Any element (X, γ) ∈ M∗1,1 is determined by the pair (, τ ),
the length and the twisting parameter of X around γ. Note that we have
φγ (, τ ) = (,  + τ ), where φγ denotes a right Dehn twist around γ. Hence
we have 
M∗1,1 ∼= {(, τ )|0 ≤  ≤ τ } (x, 0) ∼ (x, x).
The Weil-Petersson symplectic form in Fenchel-Nielsen coordinates is
given by π ∗ (ωwp ) = d ∧ dτ. Therefore, we have
   ∞ x
f ((Y )) dX = f ((Y )) dY = f (x) 1 dy dx.
M1,1 π(Y )=X M∗1,1 0 0

Integrating McShane’s identity (1.2) over M1,1 against the Weil-Petersson
volume form, we obtain
∞ ∞
 π2
Vol(M1,1 ) = 2  f () d = 2 d = .
1 + e 6
0 0

Calculation of Vg,n . To carry out a similar analysis for Mg,n we will:

4

. y) ∼ 1 + ey as x → 0. y. we want to find a function defined on Teichmüller space such that the sum of its values over the elements of each orbit of Modg. 5 . Roughly speaking.n is a constant independent of the orbit. therefore our formula for genus one hyperbolic surfaces with one geodesic boundary component (1. On the other hand. . As a special case. Ln . . we get  D(L. In the formula above. α2 (X)) + R(L1 . For any hyperbolic surface X with n geodesic boundary com- ponents β1 .3) {α1 . . (1. McShane [M] gives a version of formula (1. A central role in our approach to volumes of moduli spaces is played by the following result (§4 ): Theorem 1. .n (L) ob- tained in §9.4) γ where the sum is over all simple closed geodesics γ on X. γ (X)) = L1 . we have (§3) 2x D(x. . Li . We now turn into a more detailed account of two main steps of the proof: (I): Generalized McShane’s identity. for any hyperbolic surface X of genus one with one geodesic boundary component of length L.4) implies the original McShane identity (1. by regarding it as a geodesic of length 0.βn of lengths L1 . α1 (X). In our discussion. (I): Generalize McShane identity (Theorem 1. The result is a recursive formula for the volume polynomial Vg. and (II): Develop a method to integrate functions given in terms of the hy- perbolic length functions over the moduli space (§8). and the second sum is over simple closed geodesics γ bounding a pairs of pants with β1 and βi .1) for punctured Riemann surfaces of higher genus. γ (X). (1. we also allow βi to be a cusp of X. we have   n  D(L1 .3. γ (X)) = L.α2 } i=2 γ Here the first sum is over all unordered pairs of simple closed geodesics {α1 . α2 } bounding a pair of pants with β1 ..2) to arbitrary hyperbolic surfaces with geodesic boundary components (§4). In §3 we introduce two auxiliary functions D.1) when L → 0. we need a further generalization to bordered Riemann surfaces with geodesic boundary components. . R : R3+ → R+ related to the geometry of hyperbolic pairs of pants.

Working with bordered Riemann surfaces allows us to exploit the existence of commuting Hamiltonian S 1 -actions on certain coverings of the moduli space in order to integrate certain geometric functions over the moduli space of curves.n . For any function f : R+ → R+ . we associate the set Oγ = {[α] |α ∈ Modg.n . let [γ] denote the homo- topy class of γ and let γ (Y ) denote the hyperbolic length of the geodesic representative of [γ] on Y . Therefore. Mγg.n → R defined by (X.(II): Integration over the moduli space.n be a closed surface of genus g with n boundary components and Y ∈ Tg.  fγ (X) = f (α (X)).n -orbit of γ on X ∈ Mg.1 for the general case.n . we develop a method for integrating the right hand side of the identity for the lengths of simple closed geodesics (1. [α]∈Oγ defines a function fγ : Mg.n .n .n (L). Let Sg.n Mγg. we have   fγ (X) dX = f ◦ (Y ) dY.n 0 6 . and α ∈ Oγ is a geodesic on X } → Mg. To each simple closed curve γ on Sg. consider the covering space of Mg. The function f is constant on each level set of  and we have  ∞ f ◦ (Y ) dY = f (t) Vol(−1 (t)) dt. First. The hyperbolic length function descends to the func- tion.n ·γ} of homotopy classes of simple closed curves in the Modg.n On the other hand. α) = X. Mg. where π γ (X. See Theorem 8.n .n → R.n with respect to the Weil-Petersson volume form when γ is a connected simple closed curve. In §8.n = {(X.  : Mγg. Here we sketch the main idea of calculating the integral of fγ over Mg.3) over Mg. For any simple closed curve γ on Sg. η) = η (X).n π γ : Mγg. α) | X ∈ Mg.

n induced by twisting the surface along γ.n (γ) be the result of cutting the surface Sg.n . we have  ∞ −M (γ) fγ (X) dX = 2 f (t) t Vol(M(Sg. γ = t) to be the moduli space of Riemann surfaces homeomorphic to Sg. the situation is different when γ seperates off a one-handle in which case the length of the fiber at a point is in fact t/2 instead of t.1. M(Sg. We define M(Sg. Let Sg.n 0 where M (γ) = 1 if γ seperates off a one-handle. The method of symplectic reduc- tion can be used to show that Vg. that is Sg. So we expect to have Vol(−1 (t)) = t Vol(M(Sg. In a sequel.n (γ). and M (γ) = 0 otherwise. Thus Sg. 1).n (L) in terms of intersection numbers of tautological 7 . γ = t) as symplectic manifolds. We have a natural circle bundle −1 (t) ⊂ Mγg.n (γ). γ = t) We will study the S 1 -action on the level set −1 (t) ⊂ Mγg. On the other hand. we obtain a formula for Vg.n in terms of moduli spaces corresponding to simpler surfaces.n (γ) is a possibly disconnected compact surface with n + 2 boundary components. An alternative proof of Theorem 1.n (L) is a polynomial in L.n is that the decomposition of the surface along γ gives rise to a description of Mγg.5) Mg. γ = t)) dt.n (γ) ∼ = Sg.n (γ).n along γ.n (γ). By investigating these S 1 -actions in more detail in §8 we show that −1 (t)/S 1 ∼ = M(Sg. For any connected simple closed curve γ on Sg.n (γ) such that the lengths of the 2 boundary components corresponding to γ are equal to t.where the volume is taken with respect to the volume form − ∗ d on −1 (t). This leads to formulas for the integral of fγ in terms of the Weil-Petersson volumes of moduli spaces of bordered Riemann surfaces and the function f . γ = t) is equipped with the Weil-Petersson symplectic form.n (γ).n (γ).n − Uγ . The main idea for integrating over Mγg. But as we will see in §8. (1. The quotient space −1 (t)/S 1 inherits a sym- plectic form from the Weil-Petersson symplectic form.n    M(Sg. where Uγ is an open neighborhood of γ homeomorphic to γ × (0. γ = t)).

classes over Mg. I would like to thank Curt McMullen for his invaluable help and many stimulating discussions over the course of this work. we also recall some basic facts and results on hyperbolic geometry. Yau and Igor Riven for helpful discussions. Maxim Kontsevich. Penner has developed a different method for calculating the Weil-Petersson volume of the moduli spaces of curves with marked points by using decorated Teichmüller theory and calculated the Weil-Petersson volume of M1. A recursive formula for the Weil-Petersson volume of the moduli space of punctured spheres was obtained by Zograf [Zo]. We present some familiar concepts in a less familiar setting about the symplectic structure of the moduli space of bordered Riemann surfaces and the space of measured geodesic laminations.n .n (L) presented in §5 leads to a new proof of the Virasoro constraints for a point which is equivalent to the Witten-Kontsevich formula [K]. In forthcoming papers. 2 Background material In this section. Moreover. this symplectic method does not lead us to a recursive algorithm for calculating the volumes explicitly. However. The author would also like to thank the Max Planck Institute of Mathematics in Leipzig and the Institute for Studies in Theoretical Physics and Mathematics (IPM) in Tehran for their hospitality during the writing of this paper.2 [Pen].n (L) to intersection numbers of tautological line classes on Mg. Andrei Okounkov. Zograf and Manin have obtained generating functions for the Weil-Petersson volume of Mg. Chiu-Chu Melissa Liu. The Weil-Petersson volume of the moduli space of punctured Riemann surfaces arises naturally in different contexts[KMZ]. Acknowledgments. S. I am also grateful to Said Akbari.n [MaZ]. Applications. Recall that a symplectic structure on a manifold M is a non-degenerate 8 . Also. Izzet Coskun. we will study the connection of the polynomial Vg.1 (L) was also previously obtained in [NN] by finding a fundamental domain for the action of the mapping class group on Teichmüller space. We also relate the coefficients of the volume polynomial Vg.n [Mirz2]. Notes and references. Rahul Pandharipandeh.T. R. I would like to thank Greg McShane and Scott Wolpert for many helpful comments and dscussions. The algorithm for calculating Vg. The volume polynomial V1.n (L) with the length distribution of simple closed geodesics on a hyperbolic surface [Mirz1].

|A| Let A = ∂S and L = (Lα )α∈A ∈ R+ .n = Tg. When ∂S is nonempty.n . . . Ln ) = T (Sg.n (L) by changing the marking. . . . The mapping class group Modg. or the group of isotopy classes of orientation preserving self homeomorphisms of S leaving each boundary component set wise fixed. . . Two marked surfaces f : S → X and g : S → Y define the same point in T (S) if and only if f ◦ g−1 : Y → X is isotopic to a con- formal map. A point X ∈ T (S. Ln ). . . 0). L1 . The map f provides a marking on X by S. 0). denote the Teichmüller space of hyperbolic structures on Sg. .n (0. Also.n with geodesic boundary components of length L1 . .n be an oriented connected surface of genus g with n boundary com- ponents (β1 . . 9 .n (L1 . . . . βi = Li ) = Tg.n = Mg.n (L1 . we have Mg. .n = Mod(Sg. . By convention. . Here we briefly summarize the background material on Teichmüller theory of Reimann surfaces with geodesic boundary compo- nents. The n-fold wedge product 1 ω ∧ ··· ∧ ω n! never vanishes and defines a volume form on M . L) is a marked hyperbolic surface with geodesic boundary components such that for each boundary component β ∈ ∂S. . . βn ). . . consider hyperbolic Riemann surfaces homeomorphic to S with geodesic boundary components of fixed length. Let Sg.n is the moduli space of Riemann surfaces homeomorphic to Sg.n ) acts on Tg. Teichmüller Space. a boundary geodesic of length zero is a cusp and we have Tg.n (0. . . . Let Mod(S) denote the mapping class group of S. Then Tg. A point in the Teichmüller space T (S) is a complete hyperbolic surface X equipped with a diffeomorphism f : S → X. we have β (X) = Lβ . Ln )/ Modg.n . . The quotient space Mg.n with n boundary components of length βi = Li . . .closed 2-form ω ∈ Ω2 (M ).n (L) = M(Sg. Ln .

we have k Vol(M(S. Fix a system of pants decomposition of Sg. . i=1 where LAi = (Ls )s∈Ai . . i=1 The Fenchel-Nielsen coordinates. . . Th is symplectic form is called Weil- Petersson symplectic form. {α1 (X). . we have i=1 k M(S.n (L) and t ∈ R. . i=1 Twisting. In this paper. . We cut the surface along α.n (L) ∼ = RP +×R P by the map X → (αi (X). By work of Goldman [Gol]. ταi (X)). For a marked hyperbolic surface X ∈ Tg. . . For any simple closed geodesic α on X ∈ Tg. LAi ). τα1 (X). we can deform the hyperbolic structure as follows.1 (Wolpert). we are interested in calculating the volume of the moduli space with respect to the volume form induced by the Weil-Petersson symplectic form.n (L1 . . By work of Wolpert. over Teichmüller space the Weil-Petersson symplectic structure has a simple form in Fenchel-Nielsen coordinates [Wol1]. We have an isomorphism Tg. P = {αi }ki=1 . L) = M(Si . . LAi )). ταk (X)}. αk (X). where k = 6g−6+ 2n. A pants decomposition of S is a set of disjoint simple closed curves which decompose the surface into pairs of pants. . con- sists of the set of lengths of all geodesics used in the decomposition and the set of the twisting parameters used to glue the pieces. Note that when S is disconnected. the space Tg. The Weil-Petersson symplectic form. and denoted by ω or ωwp . Theorem 2. 10 . The Weil-Petersson symplectic form is given by  k ωwp = dαi ∧ dταi .n .n (L). the Fenchel-Nielsen coor- dinates associated with P. Ln ) carries a natural symplectic form invariant under the action of the mapping class group. L)) = Vol(M(Si . k For a disconnected surface S = Si such that Ai = ∂Si ⊂ ∂S.

and γk are distinct. twtα is the Hamiltonian flow of the length function. We denote this surface by Sg. be the isotopy class of a multi curve on Sg. . . where Uγ is an open set homeomorphic k to 1 (0. Splitting along a simple closed curve. β = L) 11 . the resulting continuous path in Teichmüller space is the Fenchel-Nielsen deformation of X along α. Namely. In other words.n (γ)) = {β1 . . . . . where φα ∈ Mod(Sg. . Consider the surface Sg.n (γ). . As t varies. γk ). Ln ) and x = (x1 . γ12 . Γ = x. disjoint simple closed curves. . we have twtα (X) = φα (X). . . the vector field generated by twisting around α is symplectically dual to the exact one form dα . Cutting the surface Let γ  k γ= ci γi . γi1 and γi2 on Sg. xk ) ∈ Rk+ .n . ∂(Sg. . gives rise to 2 boundary components. . let M(Sg.n − Uγ . . For t = α (X). βn } ∪ {γ11 .1). Now for Γ = (γ1 . i=1 where γ1 .n ) is a right Dehn twist about α. . . . . .turn left hand side of α in the positive direction the distance t and reglue back. L = (L1 . . Each connected component γi of γ.n (γ). which is a (possibly disconnected) surface with n + 2k boundary components and s = s(γ) connected components. β1 β1 β2 β2 β3 α β3 β4 α1 β4 α2 Figure 1.n (γ). . . 1) × γi around γ. Let us denote the new surface by twtα (X). γk2 }. . γk1 . By Wolpert’s result (Theorem 2.

β = L) ∼ = Mgi . β = L)). L) = Vol(M(Sg.1) i=1 Then in terms of the above notation. x. i=1 Symmetry group of a multi curve. i=1 where Ai = (α )α∈Ai .ni (Ai ). x.be the moduli space of hyperbolic Riemann surfaces homeomorphic to Sg. by i=1 Sym(γ) = Stab(γ)/ ∩ki=1 Stab(γi ).n .n (Γ. For any set A of homotopy classes of simple closed curves on Sg. define Stab(A) by Stab(A) = {h ∈ Modg. In fact.n (γ) such that γi = xi and βi = Li .n (γ).ni . Sym(γ). (2. define the symmetry group of γ.n . we have s M(Sg. L) = Vgi .n (Γ. Also. Ai = ∂Si ⊂ B. define Vg.n (Γ.n (γ). L) by Vg.n | h · A = A} ⊂ Modg. β. Γ = x.n (γ) = Sgi . when γ has extra symmetry .n (γ) as a union of its connected components s Sg.  k For γ = ci γi . β. and consequently we get s Vg.ni (Ai ). Γ = x. β. x. We can write Sg.

. i=1 First. 12 . | Sym(α)| = 1. . = ck . Note | Sym(γ)| = 1 for k γ= ci γi i=1 will put a non trivial condition on the ci s. For example | Sym(γ)| = k! implies that c1 = c2 = . k Stab(γi ) = Stab(γ). for any connected simple closed curve α.

2. of homotopy classes of simple closed curves in the Modg. Let C(x1 . Although there is no canonical simple closed geodesic on X ∈ Mg. L). We also allow the degenerate case in which one or more of the lengths vanish. Let [γ] denotes the homotopy class of a simple closed curve γ on Sg. and send γ1 to γ2 .n corresponding to [γ].n (γ2 ). 3.n (γ) is connected or Sg.n . Here.3 . α∈Oγ where f : R → R+ . Simple closed curves on X ∈ Mg.1 ∪ Sg1 . the characteristic function of [0. i = 1. A pair of pants is an oriented compact surface homeomorphic to S0.n → R+  X→ f (α (X)). It is easy to check that | Sym(γ1 + γ2 )| = 2 if and only either Sg. For background on hyperbolic geometry see [Bus]. Later we will be interested in the case where γ bounds a pair of pants with a boundary component of Sg.1 .n (γ) ∼ = Sg1 .n . we study functions of the form fγ : Mg. is determined by γ. 3 Geometry of pairs of pants In this section we study infinite simple geodesic rays on a hyperbolic pair of pants. by the definition | Sym(γ1 + γ2 )| = 2 if and only if Sg. In other words.n → X is a marking of X. x3 ) be the unique hyperbolic pair of pants with geodesic boundary curves (βi )3i=1 such that βi (C) = xi . Here we want the homemorphisem to fix each boundary component of ∂(Sg. Oγ is the set of [φ(γ)] where φ : Sg. L). When k = 2.n ) setwise.n (γ1 ) is homeomorphic to Sg. Let α (X) denote the hyperbolic length of α on X.n -orbit of γ on X. fγ (X) is equal to the number of elements of Oγ of length less than L on X. for f = χ[0. a surface of genus-0 with three boundary components. 13 .n . As an example. x2 . the set Oγ = {[α]| α ∈ Mod ·γ}.

x3 ) is equal to the geodesic length of the projection of β3 on β1 . z1 ) and (y2 . x2 . x3 ) by pasting two copies of the (unique) right angled geodesic hexagons with pairwise non-adjacent sides of length x1 /2. we can obtain C(x1 . complete geodesics in a pair of pants Each boundary component of C has two canonical points. x2 . we consider the universal cover of C. x3 ) to be the geodesic length of (y1 . Complete geodesics on hyperbolic a pair of pants. 2 of these geodesics meet β1 at y1 and y2 and spiral around β3 . On the other hand. So the 14 . y2 ). Note that we have σ(w1 ) = w2 . w2 y1 z1 β1 w1 β3 β2 Figure 2. See Figure 3. Then it is easy to check that x1 − R(x1 . Note that this length does not depend on the choice of the lift of C. Thus C(x1 . define D(x1 . the interval between y1 and y2 along β1 containing w1 and w2 . σ(z1 ) = z2 . x3 ) to be the sum of the geodesic length of (y1 . Definitions. and σ(y1 ) = y2 . the end points of the length minimizing geodesics connecting it to the other two boundary components. x2 . There is also a unique common simple geodesic perpendicular from β1 to itself meeting β1 perpendicularly at 2 points. Also. x3 ) admits a reflection involution σ which interchanges the two hexagons. More precisely. A hyperbolic pair of pants contains 5 complete geodesics disjoint from β2 . the interval between yi and zi containing wi on β1 . w1 and w2 . See Figure 2. x2 /2 and x3 /2 along the remaining three sides. z2 ). the other 2 meet β1 at z1 and z2 and spiral around β2 . x2 . x2 . For caclulating the function R. β3 and orthogonal to β1 . Define R(x1 .

z) = x − log y x−z . Also. y. D(x1 . x3 ) = D(x1 . Projection of a geodesic function D(x1 . x2 . 1+e 2 1+e 2 Basic properties of D and R. x2 ) = x1 + D(x1 . (3. x3 ) equals 2 times the distance between the projection of β2 and β3 on β1 . y) = x+y + x−y . x3 . x3 ). x3 ) + R(x1 . and R(x1 .1) e 2 +e 2 and   cosh( y2 ) + cosh( x+z 2 ) R(x. (3. x3 ) is twice the geodesic distance between two geodesics perpendicular to β1 spiraling around β2 and β3 .1. Equivalently in the universal cover of C. y. define H : R2 → R by 1 1 H(x. p2 p3 p1 r3 β1 β3 r2 r1 Figure 3. x2 ). x2 . x3 . one can explicitly calculate these functions and show that: Lemma 3. z) = 2 log −x y+z . x2 . x2 . The functions D and R are given by x y+z  e2 + e 2 D(x. Moreover.2) cosh( 2 ) + cosh( 2 ) 15 . x2 . It can be easily checked that the functions D and R satisfy D(x1 .

Now by apply- ing formula (3. Equation (3.   cosh(d(β1 . β2 ). (3.g. b. x3 /2 and d(β3 . Theorem 2.1 of [Bus]) in any geodesic quadrilateral with three right angles and consecutive sides of lengths a. x2 and x3 uniquely de- termine d(β1 .2.3) Let r3 p3 be the unique geodesic perpendicular to β3 and β1 . R : R+ → R+ satisfy the following equations: ∂ D(x. cosh(d(β1 . 2 Remark. β3 ). we obtain two convex right-angled geodesic hexagons with consecutive sides of lengths x1 /2.Proof. x − y)).4) ∂x and ∂ 1 R(x.1) shows that D is a function of x and y + z. This means that x1 . z). x). β3 ). x2 . (3. x2 /2. Using basic trigonometry of hyperbolic hexagons ( e. β3 )) On the other hand by cutting the pairs of pants along the shortest geodesics joining distinst boundary components. (3. z) = H(y + z. β3 )) = . The functions D.4. x + y) + H(z. sinh(d(β1 .3.5) ∂x 2 16 . β1 ). y. x2 . x3 ) = x1 − 2 arcsinh . z) = (H(z. x3 ) = x1 − log . √ On the other hand. β3 )) − 1 which implies equation 3. β3 ) + 1 R(x1 . we have   1 R(x1 . It is enough to calculate R(x. since arcsinh(z) = log(z + z 2 + 1) we have       1 1 cosh(α) cosh(α) + 1 2 arcsinh = 2 log + = log . sinh( x23 ) sinh( x21 ) See §2 of [Bus] for more details.1 in [Bus] ). d(β1 .g. Next lemma allows us to simplify integrals involving D and R : Lemma 3. y.3) to two geodesic quadrilaterals r1 r3 p1 p3 and r3 r2 p3 p2 in Figure 3. y.2. we get cosh( x22 ) + cosh( x23 ) cosh( x21 ) cosh(d(β1 . Theorem 2. Using basic trigonometry (e. d(β2 . sinh(α) sinh(α) sinh(α) cosh(α) − 1 therefore. we have Sinh(a) · Sinh(b) = 1. infinity and infinity (when one vertex is on the boundary at infinity ).

both D(x. There are constants c1 . c2 > 0 such that for any x ≤ 1 we have    D(x. z) which implies (3. z) + D(x.6) 1+e 2   1 1 2x R(x. ∂x e + e(y+z)/2 e−x/2 + e(y+z)/2 On the other hand. when x and y are fixed numbers as z → ∞. z) ∼ x z+y + z−y . Functions D and R are continuous on R3+ .1. y. z) ∼ z . y. By using Lemma 3. Equation (3. z) and R(x. x. y. Moreover.4) is a straight forward calculation from equation (3. y. x) ∼ y+z . we have: Lemma 3. one can show that D(x. z → ∞. since we have ∂ ex/2 e−x/2 D(x. z) → 0. −y. y.  x 1+e 2     R(x. y. z) ≤ x. y. z) 1  −(y+z)  −  y+z  ≤ c1 x e 2 . z) = 2 R(x. z) 1 1   − z+y +  ≤ c2 x e −z 2 . z) = x/2 + = H(y + z. when x is a fixed number as y.5). y. D(x. similarly.2 it is easy to verify that 2x D(x.Proof. z) → 0. y. z) ≤ x and 0 < R(x. y. we have R(x. using Lemma 3. (3. (3.  x z−y  1+e 2 1+e 2 Also. 17 .1). R(x. z) ∼ x H(y + z. y. z) go to zero when x → 0. As 0 < D(x. y. 2 Asymptotic behavior of D and R. y.7) 1+e 2 1+e 2 1 + e2 as x → 0.3. x).

bound a pair of pants if there exists an embedded pair of pants Σ ⊂ Sg.j be the set of isotopy classes of simple closed curves γ bounding a pairs of pants containing βi and βj . we generalize McShane’s identity for bordered hyperbolic Riemann surfaces with geodesic boundary components.α2 }∈F1 i=2 γ∈F1.n such that ∂Σ = {α1 . α2 ∈ ∂(Sg.7).1) {α1 . . Li .3 motivates the following definitions.1) tend to zero and β1 tends to a puncture p1 . α2 } bounding a pairs of pants with βi such that α1 .4 Generalized McShane identity for bordered sur- faces In this section. (4.n . The statement of Theorem 1. .6) and (3. . . Ln ) with 3g − 3 + n > 0. defined in the preceding section.i We will use the properties of functions D.α2 }∈F1 i=2 γ∈F1.2 : 18 . For any X ∈ Tg. let Fi be the set of unordered pairs of isotopy classes of simple closed curves {α1 . let Fi.2 (Generalized McShane identity for bordered sur- faces). α3 }. α2 . McShane [M]: Theorem 4.1.n ). • For 1 ≤ i = j ≤ n. We say three isotopy classes of connected simple closed curves. Here αi can be a boundary component and we consider punctures as simple closed geodesics of length 0. Using (3. Then we have  1  n  1 1 α1 (X)+α2 (X) + γ (X) = . α3 ) on Sg. and the geometry of complete simple geodesics on a hy- perbolic surface to get the following result for hyperbolic bordered Riemann surfaces with geodesic boundary components: Theorem 4. • For 1 ≤ i ≤ n. (α1 .n . α1 (X). γ (X)) = L1 . Embedded Pairs of pants.n (L1 . 1+e 2 1+e 2 2 {α1 . An identity for lengths of simple closed geodesics.i Note that as L1 → 0 both sides of (4. the following Corollary is an immediate result of Theorem 4. R : R3+ → R+ . Let {pi }n1 be the set of punctures of X ∈ Tg. First we state an identity for lengths of simple closed geodesics on hyperbolic punctured surfaces due to G. α2 . α2 (X)) + R(L1 . we have   n  D(L1 .

For any X ∈ Tg.α2 }∈F1 1 + e 2 i=2 γ∈F1. 2 Later we show that: Theorem 4.Corollary 4. Let E(X) be the union of all simple complete geodesics perpendicular to all boundary components and Ei = E ∩ βi . By a result due to Birman and Series [BS].1. let γx . So µ(E ∩ Uβi ) = 0 implies that µ(Ei ) = 0. E ∩ Uβi has measure zero . . Because of the structure of the collar neighborhood we have µ(E ∩ Uβi ) = sinh r × µ(Ei ). denote the complete simple geodesic perpendicular to βi such that x ∈ γx . .3 implies Theorem 4. Given x ∈ Ei . See also [B] for a related result for the lengths of common orthogonals of two totally geodesic hypersurfaces on a hyperbolic manifold. the union of all com- plete geodesics on a closed surface has Hausdorff dimension 1. Proof.4.i 2 (4. . has measure zero. where r is the width of the collar neighborhood. Remark. 2 1+e 2 1+e 2 {α1 . L2 . Union of complete simple geodesics. . Therefore. we basically follow the proof presented in [M] almost line for line by relating the topology of the union of complete simple geodesics perpendicular to all boundary components to the global behavior of simple closed geodesics.2) Notice that corollary 4. the geodesic emanating from x. Theorem 4.n (0. Ln ) with 3g − 3 + n > 0. 19 . we have   1 n 1 1 1 1 α1 (X)+α2 (X) + γ (X)+Li + γ (X)−Li = . Each Ei is homeomorphic to the Cantor set union countably many isolated points.3.2. Doubling the bordered surface along its boundary components shows that the same state- ment holds for a bordered surface. To prove Theorem 4.5. defined as above. The set Ei ⊂ βi . That is µ(E) = 0. where Uβi is the collar neighborhood around βi .

2. there exists a unique embedded pair of pants containing γ and these (not necessarily distinct) boundary components. 1. The other end spirals into a compact minimal lamination inside the surface. which will be denoted by Ω(γx ). γx falls into exactly one of the following two classes. b) If Ω(γx ) is a not a simple closed curve then x is neither a boundary nor an isolated point in Ei . In this part we give a characterization of boundary and isolated points in Ei . Note that for x ∈ Ei . for any γ joining two boundary compo- nents. Ω(γ) is actually a minimal lamination. c) The point x is a boundary point of Ei if Ω(γx ) is a simple closed curve inside the surface. the corresponding simple geodesic ray.6. The other end also approaches a (not necessarily distinct) boundary component βi in which case either the ray γx meets βi perpendicularly or spirals around it. there exists a unique simple geodesic joining them perpen- dicularly. Also in each pair of pants containing two (not necessarily distinct) bound- ary components. For any x ∈ Ei . We say a lamination γ spirals to a lamination Ω(γ) iff Ω(γ) is in the closure of γ. exactly one of the following holds: a) The point x is an isolated point of Ei if the other end of γx approaches a boundary component. 20 . Figure 4. Finding the pair of pants containing a simple geodesic Notice that. as shown in Figure 4.Characterization of boundary and isolated points in Ei . It can be easily checked that when γ is a ray. We will prove the following classification of points in Ei in terms of the behavior of the corresponding complete simple geodesics (See [M]) : Theorem 4.

There are precisely 4 infinite geodesic rays in Σ meeting β1 perpendicularly at one point (as in Figure 1).Proof of Theorem 4. let x = x1 ∈ X1 be such that the other end of γx goes up to β1 and let x2 ∈ β1 be such that γx ∩ β1 = {x. if γz is a simple geodesic ray and z ∈ {x1 . joining r1 and r2 . On the other hand. Let γyi and γzi be the ones spiraling around αi for i = 1. η2 ) be the piecewise geodesic path going from z̃ to h along γ̃z and from h to r1 (resp. z1 ) and x2 ∈ X1 ∩ (y2 . 2 such that x1 ∈ X1 ∩ (y1 . we can assume that γz meet α1 first. two geodesics perpendicular to β̃ passing through the two end points of α̃1 . x2 . z2 ) = x2 . since α̃1 is the outermost lift of α1 meeting γz the projection of η1 and η2 are disjoint 21 . x2 }. let α̃1 . Also.6(a). be the outermost lift of α1 meeting γ̃z . Universal cover of a pair of pants Σ and hence meet α1 ∪α2 . And let η1 (resp. w1 . y2 . One can easily modify the argument for other cases. α1 . and X1 ∩ (y2 . Let Σ denote the pair of pants containing γx1 such that ∂Σ = {β1 . r2 ) along α̃1 As both α1 and γz are simple. Without loss of generality. α2 }. z1 ) = x1 . Consider ψ1 and ψ2 . let β̃. γz must leave z̃ α˜1 β̃ h r2 r1 s1 Figure 5. w2 }. joining s1 and ∞. We claim that X1 ∩ (y1 . z2 ). as shown in Figure 5. y1 . In the universal cover of this pair of pants. be a lift of β1 . the projection of η1 and η2 are simple rays on the surface.

we show how one can approximate γx with simple complete geodesics using quasi geodesics: I): Good geodesic segments. θ) → (∞. z1 ].7. 0) the distance from α to its straightening tends to zero. Also. A polygon path α of segments of length at least L and bends at most θ < π is a quasi-geodesic when L is long enough compared to θ. • The arc c is almost perpendicular to α. This shows that the projections of ψ1 and ψ2 are complete simple geodesics on Σ. A path α(t) in H. To prove part (b) we construct a sequence {xj } ⊂ Ei getting close to x from both side on βi . parameterized by arclength. the projections are infinite simple geodesic rays on Σ. Let α(t) be the arc length parameterization of a simple geodesic segment on X. that is π |∠ (c (0). Therefore a1 and a2 are actually pre images of z1 and y1 . we have z ∈ [y1 . Also as (L. η1 (resp. 1] → X be a differentiable arc transverse to α such that c(0) = α(t0 ) . η2 ). Therefore. c(1) = α(t1 ). See [CEG] for more details. Since both ψ1 and ψ2 are asymptotic to a lift of α1 . 2 Next.from both α1 and α2 . One can show that any quasi geodesic is a bounded distance from a unique geodesic. η2 ) is homotopic to ψ2 ( resp. So roughly. for any x ∈ [a1 . their images spiral to α1 . 2 22 . a2 ] the curve γx meets α. Therefore. we construct paths with uniformly bounded small curvature approximating a complete simple geodesic. Quasi-geodesics. In the next 3 parts. Furthermore. we need to approximate γx with simple complete geodesics from both sides. Later. assume that γx spirals into a compact minimal lamination Ω(γx ) which is not a simple closed curve. c) is an -good geodesic arc iff • (c) ≤ . t1 . The main point is that it is easier to construct quasi geodesics approximating a complete geodesic. α (t0 )) − | ≤ . is a quasi geodesic if d(α(s). Lemma 4. t0 .  > 0 and c : [0. t0 < t1 ∈ R. We say that (α. α(t)) > |s − t| for all s and t.

c) is a positive pair the two tangent vectors to α at α(t1 ) and α(t2 ) are almost parallel. where Vc (v) is the parallel transport of vector v along c. η. t0 . We say (α. t0 . 2 and • The arc c meets the geodesic arc α in only two points. t1 . and let t0 = inf{t | γx (t) ∈ c[0. which starts at x and goes along γ(t) for t ≤ t0 then spirals around ψ(α. Let (α. (α (t1 ). t1 . α (t1 )) − | ≤ . 1] = ∅. that is we have  Vc (αt1 ) − αt0 ≤ . α(t1 )}. c (1)) agree. 1]}. Then we construct a complete simple curve. Consider the vectors α (t0 ) and c (0)) at point c(0) = α(t0 ). t0 . c) is positive (negative) if the orientation of the pairs (α (t0 ). c) be an -good geodesic segment such that α ∩ γx = ∅ . So if (α. the simple 23 . and α (t1 ) and c (1)) at the point c(1) = α(t1 ). A good geodesic segment π |∠ (c (1). γx ∩ c[0. Note that positivity only depends on the image of α and is indepen- dent of the parameterization. t0 . c). t1 . c (0)). α(t0 ) c α(t1 ) α Figure 6. t1 . c ∩ {α(t)|t0 ≤ t ≤ t1 } = {α(t0 ). II): Complete simple geodesics.

L > 0 such that if (α. t1 . c(0) = x0 be a small enough transverse arc such that for φ(a). t0 . For any . let φy denote the arc length parameterization of the leaf of λ such that φy (0) = y.good geodesic segment and L ≤ t1 − t0 . λ ∩ c is uncountable with only countably many boundary points. +r] → X . s. In this part. Note that λ is a minimal lamination. t1 . if η̃ denote its geodesic representative and y = η̃ ∩ βi . Without loss of generality. r] we have |a − b| > L or a = b. we can assume that the orientation of the pair (φ0 . then η is a simple quasi geodesic. and it is not a simple closed curve. y lies on the right(left) side of x if and only if (α. by possibly changing the direction of α. such that (φy . x) < . Let φ = φx0 and c : [−r. t. r > 0. and we have L ≤ |s − t|. and then goes back to α(t1 ) along α. φ(b) ∈ c[−r.9. c) is a δ. Also. Assume that c ∩ {α(t)|0 ≤ t < t1 } = ∅. c0 ) 24 . we want to find good geodesic segments in a non- trivial minimal lamination Ω(γx ) = λ in order to construct complete simple geodesics. φ(s) and φ(t) are not boundary points in λ ∩ c. Sketch of the proof. c) is positive(negative). More precisely. In fact. η will be a quasi-geodesic and consequently lies within a bounded distance of a unique complete simple geodesic. c) is a positive -good geodesic segment. L > 0 there exist 0 ≤ s < t and a transverse arc c and y ∈ c ∩ λ. we have: Lemma 4. Furthermore. Hence. For any  > 0 there exist δ.closed curve which goes along c from α(t1 ) to α(t0 ).8. III): Good geodesic sub-segments in a minimal geodesic lamina- tion. Given y ∈ λ. t0 . Therefore one can choose x0 ∈ λ ∩ c so that φx0 ∩ c does not contain any boundary points of λ ∩ c. Take a transverse almost perpendicular arc c such that λ ∩ c = ∅. We will use this lemma to approximate γx with complete simple geodesics. Then we have: Lemma 4. then d(y.

t1 . 0. Finding good-geodesic segments agrees with the orientation of X. 0. we have min{|t1 − t2 |. Let (α. c) and (φ.8 the result is immediate. Now define t1 . As γx spirals to λ. Similarly. The main idea is to apply Lemma 4. Let t1 = inf{ t > 0|φ(t) ∈ c[−r. Then as in Figure 7 at least one of (φ. 0) }. 2 Now we can prove part b that if Ω(γx ) is a non simple closed curve then x is not a boundary point. 2 we can define t2 = inf{ t > t1 |φ(t) ∈ c(x1 . t2 . Then from Lemma 4. t1 . and α(t1 ) is not a boundary point of λ ∩ c. 0) }. t2 . t1 .9 such that α(ti ) = c(ri ).6. and φ(t1 ) = c(x1 ). r2 ]}. (φ. as φ(ti ) is not a boundary point for i = 1. c) is a positive -good geodesic segment. t1 . c) be an -good geodesic segment in λ constructed in Lemma 4. Proof of part (b) and (c) of Theorem 4. Using the same method. φx0 φx0 0 φx0 0 0 2 c 2 c c 1 1 1 c) a) b) Figure 7. γx ∩ c[r1 . Let t0 = inf{t |γx (t) ∈ c[r1 . one can find a sequence of complete simple geodesics approximating γx from one side if Ω(γx ) is a simple closed curve inside the 25 . t2 } ≥ L. r2 ] is non-empty. c).9 to find positive -good geodesic segments inside λ and use it to construct complete simple geodesics. Also. t2 as follows. t2 .

Let x ∈ Ei such that the ray γx spirals into a simple closed geodesic α1 . βj and Ω(γah ) = Ω(γbh ) bound an embedded pair of pants inside the surface. 2 Now. Now if α is not a boundary component of X. bh ).2. That is any perfect totally disconnected compact metric space is homeomorphic to the Cantor set. bh are both boundary points of Ei . h 26 . Recall that we have a topological characterization of the Cantor set. The result follows since non-isolated points of Ei form a compact totally disconnected perfect subset of βi . x will be a boundary point of Ei .4 to prove the main result of this section. otherwise we could find y ∈ (ah . corresponding to the simple geodesics joining boundary components.6.surface in which case. In other words. then γbh ⊂ Σh which means that β1 . let Σh be the unique pairs of pants containing γah such that ∂(Σ) = {β1 . We find a natural one to one correspondence between embedded pairs of pants containing β1 and complementary intervals of Ei − Ii as follows. 2 Connection with embedded pairs of pants. Proof of Theorem 4. Let Ii be the set of isolated points in Ei . apart from countably many points. So α = Ω(γbh ) which means that Ω(γah ). For any h. Proof of Theorem 4. Similarly if α = βj is a boundary component. by the proof of part (a). Then there is a unique embedded pair of pants Σx on X such that γx ⊂ Σx . Ω(γah ).5. α}. We will use this fact and Lemma 4.6. γbh ⊂ Σh . bh ). h where ah . points in Ei are limit points. Then we can write Ii ∪ (βi − Ei ) = (ah . By Theorem 4. Let Ii ∪ (βi − Ei ) = (ah . bh ) and γy ⊂ Σh spirals into α. we can show that the set of non-isolated points of Ei is topologically homeomorphic to the Cantor set. there exists a unique simple closed geodesic α2 bounding a pair of pants Σx with β1 and α1 such that γx ⊂ Σx . Ω(γbh ) and β1 bound a pair of pants. then by Theorem 4.

(4.5) to rewrite (4. (4. .4 we have:  Li = βi (X) = |bh − ah |. bh ∈ βi . β (X)) = |ah − bh |. . where {a1 . The function Vg. (4. α1 (X).where ah . For each 1 ≤ h one of the following holds: 1. an ). . . . γ (X)) = |ah − bh |. βj and βi bound a pair of pants in X. By the definition the functions D and R in §3 in the first case we have R(Li . we have: 1 D(Li . .4) and (4. α2 } bounding a pairs of pants with βi in X and the second some is over γ bounding a pair of pants with βi and βj . Hence for any set A of positive numbers with |A| = n. the Weil-Petersson volume of Mg. . Ln .n (L1 .n (L1 . . 2 5 Statement of the recursive formula for volumes In this section we state a recursive formula for Vg. Then by Theorem 4. . 2. Lj . The proof is given later in §9. . . The two curves α = Ω(γah ) and β = Ω(γbh ) are distinct and bound a pair of pants containing βi .4) and in the second case. There exists j such that γ = Ω(γah ) = Ω(γbh ). an } = A.3) as   Li (X) = D(Li . Ln ) is a symmetric function in L1 . . . The volume function Vg. {α1 . . . Ln ) for any g and n (2g − 2 + n > 0) is determined recursively as follows : 27 .α2 } j γ j=i where the first sum is over unordered {α1 . Lj . α (X). . Statement of the recursive formula. γ (X)).n (A) by Vg. . .n (A) = Vg. we can define Vg. α2 (X)) + R(Li .n (L).n (a1 .5) 2 Now we can use (4. .3) h where |ah − bh | is the geodesic distance between ah and bh along βi .n (L). .

(5. (5.n (x.n (L1 . 1). set V0. (5.m (L)’s with 3h + m < 3g + n as follows.  = (L2 .3 (L1 . To do this.n (L1 . . L1 . the volume Vg.n (L1 . (0.1) ∂L1 where the functions ∞ ∞  1 Acon g. and Bg. we need the function H : R → R+ defined in §3 by 1 1 H(x.n (x. and L21 π 2 V1. Ln ). • For any L1 .n (x. L).1 (L1 ) = + . L) + Bg. 1+e 2 1+e 2 28 .n : Rn+1 + → R+ .n (L) = Acon g. L3 ) con- sists of only one point. 24 6 The first equation holds since the moduli space M0.n (L)) is inductively determined by : ∂    L1 Vg. L1 . L2 .n : R+ n+2 → R+ . n) = (1. Adcon g. For the calculation of V1. L) + Ag.n (L1 . We define the functions Acon g. When (g.2) 2 0 0 ∞ ∞  1 Adcon g.n (L) = • Let L Vol(Mg. L3 ≥ 0. L) x · B  g. L3 ) = 1. .n : R+ n+2 → R+ . L) = ( x y Adcon  g.1 (L) see §6. .n (L1 . L1 . L) dx dy).n (L1 . L) = ( x y Acon  g. L) dx dy). L) dx. L2 .4) 0 are defined in terms of the Vh. dcon (5. y. y) = x+y + x−y . 3). y. L2 .3 (L1 . .3) 2 0 0 and ∞  = Bg.

Ljk ). Now define Ag. n} and 0 ≤ g1 . L) is well defined. y. . 2. Vg−1. I1 ). (g2 . . g2 ≥ 0 and n1 = |I1 |.n 29 .n (x. . dcon  II) : Definition of A g.n : R+ n+2 → R+ by 1 Acon g. 2 ≤ 2g2 + n2 .1 (L) has a symmetry of order 2. . L) 2m(g−1. n) = 0 unless g = 1 and n = 1. L1 ). jk }. n} = I1  I2 . . define LI by LI = (Lj1 . . LI1 ) × Vg2 .n+1) See Figure 8(b). given L = (L1 .n . I2 ⊂ {2. I2 )) where I1 . I1 ). I) : Definition of Acon con g.n : R+ → R+ by dcon n+2  Adcon  g. . . y. where I = {j1 . L) a∈Ig. the function V (a. let  = Vg1 .n2 +1) As we will see later. The two sets I1 and I2 are disjoint and {2.n be the set of ordered paris a = ((g1 . . . L) =  · H(x + y. For notational convenience. Let Ig. . Ln ) and I ⊂ {1.n (x. The numbers g1 . the reason we have to divide by 2 in this case is that every X ∈ M1.n (L) is symmetric. . . . n) = δ(g − 1) × δ(n − 1).  Note that as the function Vg.n . L) 2m(g1 . . L1 . Now for each a = ((g1 .n1 +1) 2m(g2 . . (g2 .n+1 (x. y.Also as before. Ln ) =  · H(x + y. n2 = |I2 | satisfy 2 ≤ 2g1 + n2 . x. .n2 +1 (y. V (a. . and g1 + g2 = g. let m(g. . . y. V (a. 3. y. . L1 ). g2 ≤ g such that the followings hold: 1. n} with |I| = k. . L1 . . . y. x. . . So m(g. Define Ag. .n . x. I2 )) ∈ Ig. . LI2 ) .n1 +1 (x.

The functions Acon g. . III) : Definition of Bg.n . . L)  = 2m(g. Also. Connection with topology of the set of pairs of pants.n (x.n . .1 (L). L1 − Lj )) · Vg. In fact. L1 + Lj ) + H(x. . Remark. L2 . The function Vg.5) 2 j=2 See Figure 8(c). . In this case. using Theorem 4.n (L) = Cα · L 2α . L1 .1) is a recursive formula for calculating Vg.n .n (L)’s through an example when g = n = 1.n (L) is a polynomial in L1 . α |α|≤3g−3+n where Cα > 0 lies in π 6g−6+2n−|2 α| · Q. .1.n (L). define Bg. We will also calculate the leading coefficients of V0.n and Bg. Calculation of V1. namely:  Vg.n . the second condition in the definition of Ig.n are determined by the func- dcon tions {Vi.n−1 (x. this formula gives us the volume of Mg. Before proving Theorem 6. Ag.n (L) in terms of volumes of moduli spaces of Riemann surfaces that we get by removing a pair of pants containing at least one boundary component of Sg. . Ln ).n is equivalent to the condition that both complementary regions of the pair of pants have negative Euler characteristics. Ln . (5. See §9 for more details. In §6 we will simplify this recursive formula and use it to prove that Vg. Therefore equation (5.1.1).n (L). Finally. .j } where 3i + j < 3g + n. as in Figure 8.See Figure 8(a).1 has been described in purely combinatorial terms. it is closely related to the topology of different types of pairs of pants in Sg.n (L) is a polynomial in L (Theorem 1. . .n : Rn+1 + → R+ by 1 Bg. . L̂j . .n−1)  n 1 (H(x. 6 Polynomial behavior of the Weil-Petersson vol- ume In this section we use the recursive formula for the volumes of moduli spaces stated in §5 to establish the following result: Theorem 6. we elaborate the main idea of the calculation of the Vg. Although the recursive formula 5.2 for a hyperbolic surface of genus one with 30 .

as in the calculation of Vol(M1. V1. This result agrees with the result obtained in [NN]. we get: ∞ L · V1.one geodesic boundary component implies that for any X ∈ T (S1.2. 1+ey 1 + e−y Therefore.1) 24 6 Remark. 0 So we have ∞ ∂ 1 1 L · V1. ∂L 1+e 2x− 1 + ex+ 2 Integrating over M1.1 .1 ) in the Intro- duction. x. we have ∂ 1 1 D(L. we have  D(L.1 .1 (L) = x·( + ) dx. γ (X). 6 1 + ey 1 + e−y 6 8 0 Since we have 1 1 + = 1. by Lemma 3. Also. It seems straightforward to generalize our calculation for hyperbolic surfaces with finitely many cone singularities [NN]. we have: L2 π 2 + . ∂L x+ L L 1+e 2 1 + ex− 2 0 By setting y1 = x + L/2 and y2 = x − L/2. x) = L + L .1 (L) = x D(L. γ (X)) = L.1 (L) = (6. γ where the sum is over all simple closed curves γ on S1.1 (L). x) dx. x. we get ∞ ∞ ∞ 1 1 y1 − L/2 y2 + L/2 x·( + ) dx = dy1 + dy2 = x+ L x− L 1 + ey1 1 + ey2 1+e 2 1+e 2 0 L/2 −L/2 ∞ L/2 −L/2  y y − L/2 y + L/2 =2 dy + dy + dy = 1 + ey 1 + ey 1 + ey 0 0 0 L/2 π2 1 1 π 2 L2 + (y − L/2)( + ) dy = + . 31 . L).

0 We easily find in the following that ∞ ∞ (2i + 1)! · (2j + 1)! x2i+1 ·y 2j+1 ·H(x+y.1 ∂ L1 Vg. xn ] if and only if Pi (F ) ∈ R[x1 . Ag.n (L) is actually a polynomial in L. we get ∞ ∞ ∞ Z 2i+1 x ·y 2j+1 ·H(x+y. (m + n + 1)! 0 Now we can simplify the left hand side of the equation 6.. as follows. note that for any m.1 looks too complicated to be useful. At first glance the equation in 5. By setting Z = x + y. but by using the following elementary lemmas we will be able to simplify both sides of the equation 5. For any differentiable function F : Rn → R.2) (2i + 2j + 3)! 0 0 To prove equation 6. The following easy observation shows that it suffices to prove that Acon g. dcon ∂L1 and prove that Vg... t) dy dZ = 0 0 0 0 32 . we have T m! n! y m (T − y)n dy = T m+n+1 .2.2. (6.2.n (L) are polynomials in L.n (L) = Acon g.n (L) and Bg.. t) dx dy = F2i+2j+3 (t). dcon Lemma 6. .n (L) + Ag. t) dx.n (L) + Bg. xn )). xn ] . Definition. . For i ∈ N. t) dx dy = (Z−y)2i+1 ·y 2j+1 ·H(Z..n (L).. and we have F ∈ R[x1 . ∂xi Then Pi (F ) determines F . n ∈ N.n (L). define P (F ) by: ∂ Pi (F ) = (xi F (x1 . .n . define F2i+1 : R+ → R+ by ∞ F2k+1 (t) = x2k+1 · H(x.Polynomial behavior of H and Vg..

Simplifying F2k+1 (t).3) and (5. Lemma 6.4) we can express the functions Adcong. The following lemma helps us to proceed to the calculation of Vg. us- ing equations (5.n (L). In fact.n (L) and B g. (2i + 2j + 3)! 0 These functions play a key role in the calculation of Vg. exactly as in the calculation of V1. Acon g. ∞ (2i + 1)! · (2j + 1)! = Z 2i+2j+3 H(Z. (5. 2k + 2 2i + 1 1 + ex i=0 0 m2k+2 which is a polynomial in t2 whose leading term is 2k+2 . we have ∞ 1 1 x2k+1 · ( x+t + ) dx = 1+e 1 + ex−t 0 ∞ t (x + t)2k+1 + (x − t)2k+1 (x − t)2k+1 (−x + t)2k+1 = ( )dx + − + dx 1 + ex 1 + ex 1 + e−x 0 0  k   ∞ 2i+1 t2k+2 2k + 1 x = + 2t 2k−2i · · dx. Since ζ(0) = −1/2.n (L) in terms of the F2k+1 (t)’s and the volumes of moduli spaces of simpler Riemann surfaces.3. therefore the leading coefficient of F2k+1 (t) is t2k+2 /(2k + 2).1 (L) in the beginning of this section.2). we have F2k+1 (t)  k+1 t2k+2−2i = ζ(2i) (22i+1 − 4) .n (L). Remark.n (L). For any k ≥ 0. So the equality ∞ x2i+1 2 dx = ζ(2i + 2) (2i + 1)! (2 − 2−2i ) 1 + ex 0 implies that F2k+1 (2t) t2k+2  t2k−2i k = 22k+2 ( + · ζ(2i + 2) (2 − 2−2i )) (2k + 1)! (2k + 2)! (2k − 2i)! i=0 33 . Proof. (2k + 1)! (2k + 2 − 2i)! i=0 Therefore F2k+1 (t) is a polynomial in t2 of degree k + 1 such that the coef- ficient of m2k+2−2i lies in π 2i · Q>0 . t) dZ.

n is similar. Using equation 5. We prove that Ag. . let n (α1 . .n (L). L1 − Lj )) dx 0 is a polynomial in L1 and Lj which is immediate from Lemma 6. . αn ) be the coefficient of L2α 1 1 · · · L2α n n in the polynomial Vg. 2 7 Leading coefficients of volume polynomials In this section.1. By the induction hypothesis Vg. LTn −{1.n (L) dcon and Bg. 34 . Ln ). . .j} ) is a polynomial in x2 and {Lk }k∈Tn − {1. αn )g = (2g + n − 2) (α1 .1. . The proof is by induction on 3g + n. then we have i=1 (1. the coefficient of L2α in Acon (L) and Adcon g. Also. the proof for Adcong.n (L) is a polyno- mial in L.n (L) equals zero and we have:  n Theorem 7. i=1  n The recursive formula §5 simplifies when 2αi = 6g − 6 + 2n in which i=1 case we get a recursive formula in terms of the leading coefficients of {Vh.n (L).n (L) are polynomials in L. . i=1  n where |α| = αi .which implies the result. . . Definition. and Lemma 6. . Proof of Theorem 6.2. we find a recursive method for calculating the coefficients of Vg. .n and Bg. Also. Ag. α1 . αn )g . . . αn )g = Cg (α) × αi ! × 2|α| .n (L). . to complete the proof we have to show that ∞ x2i+1 · (H(x. . . .n−1 (x. 2 Now we can use the preceding lemma to prove that Vg.n (L) is a polynomial con in L. if one of the αi ’s is 0 or 1.1. . Let Tn = {1. . . L1 + Lj ) + H(x. n} and L = (L1 . Therefore. j}. .n (L) and calculate the leading coefficients of V0. it suffices to prove that Acon g. 2.m } with 3h − m < 3g − n. . . Let Cg (α1 . If αi = 3g − 3 + n. .3.

.n (L) and Bg. . αn 35 . L1 + Lj ) + H(x. These two recursive formulas are actually enough for determining (α1 . 2 We omit the proof of Theorem 7. It can be easily verified that the coefficient of L21 · · · L2n+1 αn in  n+1 ∞ 1 x · (H(x. When αi = n − 3. αn )g .n−1 (x. by induction on n it can be easily proved that:  n Corollary 7. · · · . . 2 αi + 2 α! × 2 3g−3+n 2 α! × 23g−3+n j=1 On the other hand. . . αn )g = 3 (2g + n − 2)(α1 . α1 .n (L). .1 equals 3 (1. . . . α1 . α1 . αn )g 2 · = × . . αn )g = (α1 . We prove the Theorem by induction on n.3. . .j} ) dx 2 j=2 0 equals 2 αj +2  n (α1 . LT −{1. . αn )0 = . α1 . . . . αn )g 3 (2g − 2 + n) (α1 . the leading coefficient of F2i+1 (m) equals 1/2i + 2.1.n+1 (L) by using the recursive formula for the volume polynomials. .  n Theorem 7. αi =0 Proof of Theorem 7. αn )g which implies the result. . .2. . . . . αn )g . In this case. . . . αn )g when g = 0. . . As in the proof of Lemma 6.2 since it is quite similar. L1 − Lj )) · Vg. then we have i=1  (0.3. . The coefficient of L21 · · · Ln+1 2 αn on the left hand side of equation 5. . . To do this. . we calculate the coefficient of L21 · · · Ln+1 2 αn in Vg. . . α! × 23g−3+n+1 where α! = α1 ! · · · αn !. . we have 3 (1. If αi = 3g − 2 + n. . we have i=1   α1 + · · · + αn (α1 . there is no L2α 1 · · · Ln+1 term in Ag. . αi − 1. . . . 1 2αn dcon Therefore.

and i=1 M (γ) = |{i|γi seperates off a one-handle from Sg. [α]∈Mod ·[γ]  k where α (X) = ci γi (X). γk ). For any γ = ci γi . . αn )g = ψ1α1 · · · ψnαn . .n .n (L) x∈Rk+  k where |x| = ci xi . .n (L). . .n [Mirz2].n where ψi denotes the Chern class of the ith tautological line bundle over Mg. Equations in Theorem 7.  k Theorem 8. i=1 where ci > 0 and γ1 . defines a function fγ : Mg. In a sequel we prove that  (α1 .n (L) with i=1 respect to the Weil-Petersson volume form is given by   2−M (γ) fγ (X) dX = f (|x|) Vg.Remark. 8 Integration over the moduli space In this section. | Sym(γ)| Mg. i=1 In this section we establish the following result for integrating the func- tion fγ over Mg. .1. γk are disjoint non homotopic simple closed curves on Sg.2 are reminiscent of the dilaton and string equations for the intersection pairings over the moduli spaces [Har].n (Γ.1 and Theorem 7.n (L) → R. .  fγ (X) = f (α (X)). x.n }| 36 . . β. Mg. let Γ = (γ1 . L) x · dx dt. For  k γ= ci γi . the integral of fγ over Mg. .n (L). . we investigate the Weil-Petersson symplectic structure of Mg. . For any f : R+ → R+ . .

(8. β = L)). x. and for any x = (x1 . More precisely.n (L). See [McD] for more details. z) is in fact a function of x and y + z (§3).n (Γ. Since later will use lemma 8.n (Γ. Here x·dx = x1 · · · xn ·dx1 ∧· · ·∧dxn .1 over Mg. Remark 2.1 to integrate the left hand side of equation 4. ψ1 = id. . Here we are interested in the case where XH generates an S 1 action on M .1 integrating fγ . Γ = x.ni (Ai ). it is essential that D(x. By Theorem 8. Vg. In fact | Sym(γ)| = 1 will put a non trivial restriction on the ci ’s. .) = dH(.n (γ) = Si . L) is given by Vol(M(Sg. In other words. if g(γi ) = γj for g ∈ Sym(γ) then ci = cj . i=1 Remark 1. even for a compact Riemann surface. β. We can write Sg. the vector field XH determined by ω(XH . Integration and covering.n (γ).) is called the Hamiltonian vector field associated to H. Hamiltonian circle actions. L) = Vgi . y. Let (M. Then we have k Vg. reduces to the calculation of volumes of moduli spaces of bordered Riemann surfaces. ω) be a symplectic manifold. where ψt is the integral of the vector field XH . The Hamiltonian function H in this case is called the moment map of the action.1) i=1 where Si ∼ = Sgi . Hence we have  k  k ci γi = ci g·γi i=1 i=1 . .n (γ) as a union of its connected components s Sg. . the symmetry group of γ = ki=1 ci γi . β. x. Then for any smooth function H : M → R. The terms Sym(γ) and M (γ) appear when γ has some extra symmetry.  See §2 for the definition of Sym(γ). . Let π : X1 → X2 37 . xk ) ∈ Rk+ . and Ai = ∂Si .ni .

h · γk ).n and functions defined over them whose push forward to Mg. For h ∈ Modg.n (L)’s. let OΓ be the set of homotopy classes of elements of the set Mod ·Γ. η)| X ∈ Mg. Let π Γ : Mg. .n let h. . .n (L) be the projection map defined by π Γ (X. Let φγ ∈ Modg. Then v1 = π −1 ∗(v2 ) defines a volume form on X1 . .n (L)Γ defined by the following space of pairs: {(X. . Coverings and volume forms of the Mg. If f is in L1 (X1 . .Γ = (h · γ1 . ηi ’s are closed geodesics on X}. v2 ) and we have   (π∗ f ) dv2 = f dv1 .2) X2 X1 Next we construct coverings of Mg.be a covering and v2 a volume form on X2 . (8. then the push forward  (π∗ f )(x) = f (y) y∈π −1 {x} defines a function in L1 (X2 . Consider Mg. Then . η) = X.n (L) . As in §2.n is constant.n (L)Γ → Mg. η = (η1 . . ηk ) ∈ OΓ . v1 ). .n denote the Dehn twist along γ.

Twisting and the Weil-Petesson symplectic form.n (L)Γ = Tg.n (γ). In fact. and Mg. .n (L)/Gγ .n (L)Γ is closely related to the moduli space of hyper- bolic structures over Sg. s GΓ = Stab(γi ) ⊂ Mod(Sg.n ) i=1 is generated by the φγi ’s and elements of the mapping class group of Sg. . The results of this section will arise from the existence of k commuting Hamiltonian S 1 -actions 38 . γk }.n (L)Γ which is the same as the form π Γ∗ (wwp ). the space Mg. As the Weil-Petersson symplectic structure on Teichmüller space is invari- ant under the action of the mapping class group. . it induces a symplectic structure on Mg.n cut along {γ1 . .

We first describe the corresponding R-action on Tg. . Induced flows on Mg. Since α (X) = α (twαt (X)). . .n (L)Γ L Rk ←−−−− Mg. LAi ) i=1 sending each point X ∈ Tg.. . Then the level set Tg. .tk ) : Tg. .n (L) as defined in §2. . See §2 for more details.n to the surface that we get by cutting X along components of γ. Then φ1α = φα ∈ Stab(α) is the Dehn-Twist around α. . is a symplectomorphism.n (L)Γ  π  Mg. Now Theorem 2.n (L)Γ .n → Rk+ denote the length function defined by Γ (X) = (γ1 (X). ..n (a) preserving the Weil- Petersson symplectic form. . For any (a1 .n (L) 39 . ak )/R → k T (Si . Consider the length-normalized twist flow.on Mg. the map φγ(t1 . induced by twisting the surface along connected components of γ. . By cutting the surface along γ we get a Riemann surface with geodesic boundary components.n (a) gives rise to an action of Rk on the level set Tg. .2..n (a) = −1 Γ (a) carries a natural volume form − ∗ (dγ1 ∧ · · · dγk ). γk (X)).n (a) → Tg. Let Γ : Tg. ak ) ∈ Rk+ . The length function Γ descends to a function LΓ on Mg. the canonical map s s : −1 Γ (a1 .. given by φtα (X) = twαt·α (X) (X).n (L)Γ .1 implies the following result: Lemma 8.

and the projec- tion map π : Mg. The quotient space.n (L)Γ  carries a natural volume form va induced by − ∗ dL = − ∗ ( i dγi ).n (L)Γ∗ (a) = Mg. Note in general the twisting parameter along γi can be between 0 and γi . where twηt i (X) is obtained by cutting X along ηi .n (L)Γ∗ (a). ak ).n (a1 . a generic point in Mg. The reason is that every X ∈ M1. twisting to the right by hyperbolic length t and regluing the boundaries. In the case of a simple geodesic γi separating off a one-handle (the elliptic tail case) Stab(γj ) contains a half twist and so τ varies with fundamental region {0 ≤ τ ≤ γi /2}. Therefore. η) = (ηi (X)). has closed orbits on MΓg. inherits a symplectic structure from the symplectic structure of Mg. .n (L)Γ∗ (a).n (L)Γ (a)/T k . . .n (L)Γ . .n (L)Γ (a) = L−1 (a1 .  where T k = ki=1 S 1 . Then for ti = t (Y ) Li (Y ). we get Vol(π −1 (U )) = 2−M (γ) Vol(U ) · a1 · · · ak . . n) = (1. Mg. ak ) ⊂ Mg. twi . we have: • Each level set Mg. twi i is the Dehn twist of Y along ηi which equals Y in Mg.n (L) does not have any non trivial automorphism fixing the boundary components set wise.n (L) .1 (L) comes with an elliptic involution. The construction of the Fenchel-Nielsen flow defined on Teichmüller space is equivariant with respect to the action of the mapping class group. • The Hamiltonian flow of Li . but when (g. .Γ Therefore.n (L)Γ (a) → Mg. .n (L)Γ → Rk+ gives rise to the action of T k = S 1 × · · · × S 1 by twisting along γi proportional to its length. . 1). η) = (twηt i (X). That is we have twit (X. (8. Therefore. For an open set U ⊂ Mg. where LΓ (X. the Hamiltonian flow of L2 /2 : Mg.3) 40 . η). since − ∗ dL2 = − ∗ dL/L.

 Mg. we get Vol(L−1 (a1 . · · · . β = L.n (L)Γ∗ (a) is a symplectic quotient space ( See [Ki]). . By what we said.n (L)Γ∗ (a) ∼ π = Mgi .3. . Lemma 8. .3 and equation 8.Ai    Mg. Then the integral of Fγ over Mg. Γ = a)).n (L)Γ → R by Fγ (Y ) = F (L(Y )). Note that the function Fγ is constant on each level set of Mg. Now we can use the preceding lemma to integrate certain functions over the covering space Mg. Lβ = L) ∼ Γ∗ = Mgi .where M (γ) is the number of connected components γ separating off a one- handle.n (γ). Integrating geometric functions. For any k-tuple Γ = (γ1 . .n (L) Therefore.n (L)Γ (a) of L. we have the following result: Lemma 8. Proof. x∈Rk+ Mg. Using Lemma 8. .n (L)Γ .ni (Ai ) i=1 is a symplectomorphism.n (L)Γ where x = (x1 .n (γ). For any function F : Rk → R. In [Mirz2].ni . .n (γ). define Fγ : Mg. . and the space Mg.4. β = L. we use this fact to relate the volume polynomials to the intersection pairings of tautological classes over the moduli space. Remark.3. the canonical isomorphism s s : Mg.n (L)Γ (a) −−−−→ Mg.n (L)Γ is given by   −M (γ) Fγ (Y ) dY = 2 F (x) Vol(M(Sg.n (L) (a) → M(Sg. xn ) and x · dx = x1 · · · xn · dx1 ∧ · · · ∧ dxn . . L2 /2 is the moment map for the T k action. ak )) = 2−M (γ) a1 · · · ak · Vol(M(Sg. and as a result we get 41 . Γ = x)) x dx. Γ = a. γk ) of disjoint simple closed curves.

(x.n (L)Γ (a) F (a) · Vol(M(Sg. we have.1.n (L) equals  −M (γ) 2 f (|x|) Vol(M(Sg. γ = a)) · a1 · · · ak . using the notation of Lemma 8.3.t)∈V On the other hand. β = L)) · x · dx dt. .n (L) Mg. (8.n (L)Γ 2 Now we are ready to prove the main result of this section Proof of Theorem 8. .n (L)Γ  k where F (x1 . Now we can use Lemma 8. Γ = x.n /∩ Stab(γi ) Therefore by using equation (8.n (γ). The function π∗Γ f : Mg.4) h∈Modg.   f (g·γ (X)) = Sym(γ) · f (α (X).4 to see that i=1  π∗Γ f (X)dX Mg.4 the integral of π∗Γ f over the moduli space is given by   Γ π∗ f (X)dX = Fγ (Y )dY.1.n 42 .n (L) → R+ is given by  π∗Γ f (X) = f (h·γ (X)).n (γ).2) and Lemma 8. β = L. xk ) = f ( ci xi ). we have   Fγ (Y ) dY = I(x) dx x∈Rk+ Mg.n /∩ Stab(γi ) [α]∈[γ]·Modg. g∈Modg. since by Theorem 2. . Now the result is immediate. .  I(a) = F (γ (X)) dX = 2−M (γ) × Mg. Mg.

n (L) s stated in §5. We characterize the set of topologically different pairs of pants containing β1 . β (X)).j / Modg.n (L).1) {α1 . Li .i where as in §4. Topology of pairs of pants on a surface. for any X ∈ Tg. Lj . Hence we get π∗Γ f (X) = Sym(γ) · fγ (X). γ (X)) = L1 .2) γ∈F1. . 43 . (9. . (α1 . j=2 where D̃ and R̃j are functions defined on Mg. (9.n (L). This enables us to apply Theorem 8.2 to derive the recursive formula for the Vg.α2 }∈F1 i=2 γ∈F1. . βj }. α (X). F1 and F1.n (L).j and  D̃(X) = D(L1 . By Theorem 4.j are respectively in one to one correspondence with the set of pairs of pants containing β1 and {β1 . .1 we get  n D̃(X) + R̃j (X) = L1 .α2 )∈F1 Then from 9.2.n to reformulate R̃j and D̃ as push forwards of functions defined over certain coverings of the moduli space of the form described in §8.where Sym(γ) = | Stab(γ)/ ∩ Stab(γi )|. We use the description of Fi / Modg.n (L1 . Ln ) we have   n  D(L1 . γ (X)).n and Fi. Idea of the calculation of Vg.1 and integrate these functions over Mg. Now let  R̃j (X) = R(L1 . 2 9 Volumes of moduli spaces of bordered Riemann surfaces In this section we use the identity for lengths of simple closed geodesics in Theorem 4. α2 (X)) + R(L1 . α1 (X).

. β2 βi1 . x . . a):|∂(Σ) ∩ ∂Sg. x . β1 βin1 β1 Σ Σ βj1 y y βjn2 βn a) b) β1 β2 Σ x βj c) βn Figure 8..n | = 2 44 . separating case b): non-separating case c): |∂(Σ) ∩ ∂Sg. .n | = 1.

n+1 (See also the definition of Acon g. 1). Proof of the recursive formula. If ∂(Σ)∩∂(Sg.n (L) = R̃j (X) dX. For (g. Theorem 9. We can integrate both sides of the equation  n  D̃(X) + R̃i (X) = L1 i=2 over Mg.n (Σ) can have 1 or 2 connected components. and Sg.n (Σ) is homeomorphic to Sg. then ∂Σ ∈ F1.n ) = {β1 . and Sg.n−1 (See also the definition of Bg. I): Σ contains two boundary components.n (Σ) is homeomorphic to Sg−1.n (L) satis- fies ∂ L1 Vg. 3). nevertheless the orbits can be characterized by the topology of their complementary regions which is determined by the number of the connected components.n (L).c. as in Fig 7. II): Σ contains one boundary component. Σ is non-separating: In this case. as in Fig 7. Now we are ready to prove the recursive formula stated in §5. the elements of Ig.n (L) Mg.n ) = {β1 } then Σ ∈ F1 . Mg.n ). n) = (1.n (L) Next we calculate the integrals  Rjg.n ).n (L) with respect to the volume form induced by the Weil-Petersson symplectic form. The action of Modg.n (L) + Bg. (0.n (L) = Acon g.1.n are in one- to-one correspondence with different topological types of separating pairs of pants such that ∂(Σ) ∩ ∂(Sg.n (L).2 we get    R̃j (X) dX + D̃(X) dX = L1 · Vg. Sg.n ) = {β1 } (See also the definition of Adcon g. (9. dcon (9. Σ is separating: In this case .n (L) 45 .j .4) 2≤j Mg. the volume function Vg. as in Fig 7. Let Σ be a pair of pants such that β1 ∈ ∂(Σ). genus and the number of boundary components of each connected component.3) ∂L1 Proof.n on F1 is not transitive.n (L) + Ag.n ). If ∂(Σ) ∩ ∂(Sg.b. Then as in Figure 8 one of the following holds. βj }. Therefore from equation 9.a.

.n (γj ). x) · Vol(M(Sg. by using Theorem 8.n (L) = D̃(X) dX. Consider the map π γj : Mg. For 1 = j the mapping class group Modg.n (γj ) = Sg.n−1) =2 x · R(L1 .n ·{γ} = F1. γj (X)). Lˆj . . As Sg. .n (L).n (L)γj → Mg. γ (X)). Ln )dx. Lˆj .j / Mod and γ R̃j (X) = π∗ j Rγj .n−1) Rjg. . Therefore. γj = x. Lj . . . and | Stab(γ)| = 1. L)) dx 0 ∞ −m(g. .n equals ∞ 2−m(g.5 ∂ Rj (L) ∂L1 g. . 2 0 46 .n−1) x (H(x. Lj . L1 +Lj )) Vg. Mg.j .j we have Modg. and define Rγj : Mg.n (L)Γ → R+ by Rγj (X) = R(L1 . . Lj .n (L) =2 x · R(L1 .j . . from equation 3.j and for any γ ∈ F1. L1 −Lj )+H(x. Lj . L2 . . L2 .n acts transi- tively on F1. . x) · Vg.and  Dg.n−1 (x.n−1 . Hence.n (L) I): Integrating R̃j .n−1 . we get γ  π∗ j Rγj (X) = R(L1 . 0 which can be calculated in terms of Vg. Let γj be a simple closed curve in F1. . γ∈F1. . .n−1 (x. Ln )dx.1 to show that we have ∞ −m(g. .

. a∈Ig. . It is essential that by Lemma 3.n Choose γ ∈ Acon and also. choose αa an element of the set Aa . 47 .n2 +1 . respectively homeomorphic to Sg1 . (g2 . we can use Theo- rem 8. for each a ∈ Ig. βin1 } ⊂ ∂S1 .1. let Ig. See §2 • For a ∈ ((g1 .n ) is transitive and we have F1 = Acon Adcon . α2 (X)) is in fact a function of L1 and α (X) = α1 (X) + α2 (X). α1 (X). {βj1 . α2 } ∈ F1 . βjn2 } ⊂ ∂S2 . let α = α1 + α2 .n (L). • Define Acon to be the set of α1 + α2 such that the complement of the pair of pants containing β1 . where Adcon = Aa . For {α1 . α2 is a connected surface of genus g − 1 with n+1 boundary components.n we have  n ∂ Rj (L) = Bg. See Figure 8. .n .n be the set of all possible combinations of the genus and set of boundary components of the complementary regions of elements of F1 . C by C = {αa | a ∈ Ig.1 for α = α1 + α2 .5) ∂L1 g. let Aa be the set of α = α1 + α2 such that the complement of the pair of pants containing β1 .n j=2 II): Integrating D̃ .n orbits of F1 .n1 +1 and Sg−g1 . As in §5. (9. We can classify the orbits of the action of the mapping class group as follows. such that we have: {βi1 .n } ∪ {γ}. Here we sketch the calculation for D̃.n . from the definition of Bg. I). Define the set of representatives of the distinct orbits of Ig. The action of the mapping class group on Acon and Aa (a ∈ Ig. . . Therefore by classifying the Modg. . α1 . the function D(L1 .Hence. (See Figure 8). . α1 and α2 is a disjoint union of two surfaces S1 and S2 .n . Then | Sym(α)| = 12 . . J)) ∈ Ig.

pages 3–92.4. dcon (9. [CEG] R.6) ∂L1 Now the result is immediate from equations 9.6.4 and the definition of Acon and Adcon . and P. S. and we have:  D̃(X) = π∗α Dα (X). Series. [Bus] P. α2 (X)). A. So by considering 1/2 in equation 5. Math. 54(1984). Buser. and I1 = I2 = φ. as the sum in the definition of A g. The orthogonal spectrum of a hyperbolic manifold. α=α1 +α2 ∈C where D̃ : Mαg. Therefore. Notes on notes of Thurston. Birman and C. n is over ordered pairs ((g1 . Geodesics with bounded intersection number on surfaces are sparsely distributed. [Gol] W. [BS] J. I2 )) in fact every term in the integral appears twice ex- cept for the term corresponding to the g1 = g2 . otherwise | Sym(α)| = 1. Cambridge University Press.5 and 9. J. Birkhäuser Boston. for α ∈ Aa we have | Sym(α)| = 2 if and only if I = J = φ and g1 = g2 . Also by what we showed in §2. In Analytical and Geometric Aspects of Hyperbolic Space. 1987. The symplectic nature of fundamental groups of sur- faces. Topology 24(1985). 217–225. Also. Amer. 1139–1159.n (L) → R+ is defined by Dα (X) = D(L1 . Canary. Remark. Goldman.3. The term 1/2 in equation 5. Geometry and Spectra of Compact Riemann Surfaces. I1 ). Basmajian. 48 . Green. Epstein. (g2 . D. D.n (L). we will take care of the Sym(α). 1992.n (L) + Ag. from eqaution 3.2 comes from sym(α) when α is non dcon seperating. Adv.Hence C ∼ = F1 / Modg. we get ∂ Dg. Math. 2 References [B] A.n . α1 (X). 9. 200–225. B. 115(1993).n (L) = Acon g.

Penner. Wolpert. Harris and I. Morrison. [K] M..(2) 118(1983). 9(1998). 49 . Y.. Ann. Comm. [Wol2] S. RI. 607–632. Näätänen. 2003. Amer. J. Phys. 3241–3252. Intersection on the moduli space of curves and the matrix airy function. [Wol1] S. Math. McShane. Fourier(Grenoble) 50(2000). 2003. 1998. [Pen] R. 501–528. McDuff. Springer-Verlag. 1993. 147(1992). Math. [Mirz2] M. 115(1982). [M] G. Growth of the number of simple closed geodesics on a hyperbolic surface. Annals of Math. 35(1992). and D. 491–523. Zograf. Math. Weil-Petersson volumes. Kontsevich. 132(1998). math. Phys. Mirzakhani. Wolpert.Manin. Proc. volume 187 of Graduate Texts in Mathematics. Providence. 519–535. Preprint. 135–171. Kirwan. Math. Zograf. Introduction to symplectic topology. Differential Geom. Higher Weil-Petersson vol- umes of moduli spaces of stable n-pointed curves. [MaZ] Y. Amer.Appl. Soc. 1999. Math. Mirzakhani. Differential Geom. Math. Inst. 736–787. volume 150 of Contemp. 181(1996). [KMZ] R. Momentum maps and reduction in algebraic geometry. pages 367–372. [Mirz1] M. Simple geodesics and a series constant over Te- ichmüller space. Amer. Comm.[Har] J. Soc. The Weil-Petersson volume of the moduli space of punc- tured spheres.. In Mapping class groups and moduli spaces of Rie- mann surfaces. The Fenchel-Nielsen deformation. [NN] T. Kaufmann. Invertible cohomological field theories and Weil-Petersson volumes. Ann. 129(2001). [Ki] F. [McD] D. 559–608. Preprint. On the homology of the moduli space of stable curves. Areas of two-dimensional moduli spaces. Moduli of curves. Manin and P. Zagier. of Math. Soc. Nakanishi and M. Weil-Petersson volumes and intersection theory on the moduli space of curves. [Zo] P. Invent.