Dimension Formulas of Siegel Modular Forms

of Weight 3 and Supersingular Abelian

Surfaces(Revised version)
Tomoyoshi Ibukiyama
Osaka University

In this article, first we give dimension formulas of the space S3 (Γ) of

Siegel cusp forms of weight 3 of degree 2 belonging to discrete subgroups
Γ of parahoric type, including Hecke type groups Γ0 (p), paramodular groups
′
K(p), Klingen type groups Γ0 (p) and the Iwahori subgroup B(p) of any prime
level. As for weight k ≥ 4, the dimension formulas for Siegel modular forms
of degree two are explicitly known for many discrete subgroups and there are
two well known methods to calculate these: the Riemann-Roch-Hirzebruch
theorem for k ≥ 4 and the Selberg trace formula for k ≥ 5. Both methods
does not work for weight k = 3 in general and there are no general ways to
calculate the dimension for k ≤ 3. Before our present work, the dimensions
of S3 (Γ) were calculated only for finitely many (conjugacy classes of) discrete
subgroups by rather complicated technical numerical calculation depending
heavily on each discrete group Γ (cf. e.g. Poor and Yuen [23].) In this paper,
we use holomorphic Lefschetz formula (a group invariant version of Riemann-
Roch-Hirzebruch). The obstruction of the cohomology in this formula does
not vanish in general, but by showing that it vanishes for the above discrete
groups, we obtain our dimension formula. As for paramodular groups and
Hecke type groups, our formula coincides with those conjectured in [17] and
[7].
Secondly, we give a geometric meaning of the dimensions of weight 3. We
shall show that the dimensions of cusp forms of weight 3 is related with the
geometry of principally polarized super-singular abelian surfaces in several
ways. It was known that certain arithmetic invariants of the locus of super-
singular abelian surfaces in the moduli of principally polarized abelian sur-
faces over algebraically closed field of characteristic p are related with some
class numbers of the compact real form of the split symplectic group Sp(2, R)
of size four(due to Katsura, Oort and partly myself). These class numbers

1
are nothing but the dimension of certain automorphic forms of weight 0 of the
compact twist. In general, automorphic forms of the compact twist should
correspond with Siegel modular forms by Langlands conjecture. A precise
conjecture of this type on bijective correspondence of automorphic forms has
been formulated in [14] and [13]. We have not proved this conjecture it-
self, but by our dimension formula, we can show at least certain dimensional
equalities between these forms belonging to compact or non-compact real
forms of the symplectic group. This leads to the above interpretation.
Historically the above results on dimensions and geometric interpretation
are a part of the conjectures by the author in [17] first announced in Confer-
ence on L functions at Kyushu University. The author would like to thank
Professo Takayuki Oda for asking him the possibility to use Riemann-Roch
theorem at this talk. He also thanks Professor Ryuji Tsushima for giving
him a kind guidance to cohomological methods with which the author was
not familiar before.

1 Definition of Siegel Modular Forms

We denote by Hn the Siegel upper half space of degree n.
Hn = {Z = X + iY ∈ Mn (C); t Z = Z, Y > 0}.
The symplectic group
Sp(n, R) = {g ∈ M2n (R); gJ t g = J}
( )
0 1n
acts on Hn is the usual way, where J = . The weight of Siegel
−1n 0
modular forms is given by an irreducible rational represenstation (ρ, V ) over
C of GLn (C). We take a discrete subgroup Γ ⊂ Sp(n, R) with vol(Γ\Hn ) <
∞. A Siegel modular form of Γ of weight ρ is defined to be a V -valued
holomorphic function F of Hn such that
(F |ρ [γ])(Z) := ρ(CZ + D)−1 F (γZ) = F (Z)
( )
A B
for all γ = ∈ Γ, with the standard regularity condition at each
C D
cusp when n = 1. When ρ = detk , we say F is of weight k.
When n = 2, we write ρk,j = detk Symj , where Symj is the j-th sym-
metric tensor representation of GL(2) on Cj+1 . We denote by Mk,j (Γ) the
space of Siegel modular forms of weight ρk,j belonging to Γ. A function
F ∈ Mk,j (Γ) is called a cusp form if it vanishes on the boundaries of the
Satake compactification and the space of such forms are denoted by Sk,j (Γ).

2
2 Dimension Formulas
2.1 Some general remarks
First we give some general remarks on dimension formulas. For any degree
n, if k is big enough, there are some theoretical way to calculate dimensions,
though it is often too complicated to execute calculation. Two possible ways
to calculate dimensions are
(1) Riemann-Roch-Hirzebruch and Lefschetz fixed point theorem. Here we
need the assumption that k ≥ n + 2 for the vanishing of the obstructions of
cohomology in general.
(2) The Selberg trace formula. Here we need the assumption that k > 2n for
the convergence of the kernel function of Godement.
By the method (2) and as an application of special values of zeta functions
of prehomogeneous vector spaces, we can give an explicit conjectural dimen-
sion formula for Sk (Γ(N )) of any weight k > 2n for the congruence sub-
groups Γ(N ) with N ≥ 3 of any degree (Joint work with H. Saito, cf. [19].)
This matches the known result for n ≤ 3 by Morita, Christian, Yamazaki,
Tsushima. The condition k > 2n comes from the condition that the Selberg
trace formula is valid, but actually if this conjecture is true for k > 2n, then
it is automatically true for k > n + 1 by (1).
Now if the weight is very small, i.e. if k < n/2, all the Siegel modular
forms are so called singular modular forms and there are no cusp forms. There
exists no general way to calculate the dimensions of Sk (Γ) for n/2 + 1 ≤ k ≤
n + 1. Now we assume that n = 2. For k ≤ 3, there exists no general way
to calculate dimensions. In this article we treat the case n = 2 and k = 3
for relatively big discrete subgroups Γ of Sp(2, R). The proof is related with
some new vanishing theorem for M1,j (Γ) for j > 0, which was first obtained
for j = 0 in the joint work [20] with Skoruppa. This depends on a choice of
Γ, and we cannot expect that it vanishes for general Γ.

2.2 Parahoric subgroups

We consider discrete subgroups Γ of parahoric type here. These groups are
defined as follow. Let N be any natural number. We put
 
Z NZ Z Z
 Z Z Z Z
B(N ) = Sp(2, Z) ∩ 
N Z N Z Z Z


NZ NZ NZ Z

3
and  
0 0 0 −1
 0 0 −1 0 
ρN = 
0 N 0
,
0
N 0 0 0
We also put
 
Z NZ Z Z
 Z Z Z N −1 Z
K(N ) = Sp(2, Q) ∩   Z NZ Z

Z 
NZ NZ NZ Z
 
Z Z Z Z
 Z Z Z Z
Γ0 (N ) = Sp(2, Z) ∩ 
N Z N Z Z Z 


NZ NZ Z Z
 
Z NZ Z Z
 Z Z Z Z
Γ0 (N ) = Sp(2, Z) ∩  
′

 Z N Z Z Z
NZ NZ NZ Z
 
Z Z Z Z
 Z Z Z N −1 Z
Γ0 (N ) = Sp(2, Q) ∩   = ρ−1 Γ′0 (N )ρN
′′

N Z N Z Z Z  N

NZ NZ Z Z
The group K(N ) is called paramodular group of level N .
When N = p is a prime, a discrete group Γ such that B(p) ⊂ Γ ⊂ Sp(2, Q)
′ ′′
is either Sp(2, Z), ρp Sp(2, Z)ρ−1
p , K(p), Γ0 (p), Γ0 (p) or Γ0 (p) by virtue of the
well-known Bruhat-Tits theory. The p-adic completion of Γ is a so-called
parahoric subgroup in Sp(2, Qp ). Here we call such Γ a discrete subgroup of
parahoric type.

2.3 Dimension formulas of weight 3

In this section we assume that p is any prime. The dimension formulas for
dim Sk (Γ) where Γ is any discrete subgroup of parahoric type of prime level
p were known for k ≥ 5 (cf. [7], [14], [9]), all obtained by the Selberg trace
formula. We give new results for k = 3 in this section and results for k = 4
in the next section. Outline of the proofs of these theorems will be given in
the last section. We note that, by Freitag [4] p.155 Hilfssatz 2.1, 2.5 and Satz
2.6, cusp forms of weight 3 of Γ correspond bijectively with sections of the

4
canonical divisor of any smooth compactification Γ\H2 of Γ\H2 . Although
we do not use this fact at all in our proof, we can deduce several results on
Siegel modular varieties from this fact and our new dimension formulas. For
example, by Freitag loc. cit., Siegel modular varieties Γ\H2 are not rational
for any Γ such that dim S3 (Γ) > 0, so we can give many such examples of Γ
from below.

Theorem 2.1 For paramodular groups K(p),

we have dim S3 (K(2)) = dim S3 (K(3)) = 0, and for p ≥ 5 we have

dim S3 (K(p)) =
( ( ))
1 1 −1
−1 + (p − 1) + (p + 1) 1 −
2
2880 64 p
( ( ))
5 −1
+ (p − 1) 1 +
192 p
( ( ))
1 −3
+ (p + 1) 1 −
72 p
( ( ))
1 −3
+ (p − 1) 1 +
36 p

 2/5 if p ≡ 2, 3 mod 5
+ 1/5 if p = 5

0 otherwise
( ( )) {
1 2 1/6 if p ≡ 5 mod 12
+ 1− +
8 p 0 otherwise

Numerical examples.

p 2 3 5 7 11 13 17 19 23 29 31 37
dim 0 0 0 0 0 1 1 1 1 2 2 4

5
Theorem 2.2 For Hecke type groups Γ0 (p), we have S3 (Γ0 (2)) = S3 (Γ0 (3)) =
0, and for p ≥ 5 we have

dim S3 (Γ0 (p)) =

(p + 1)(p2 + 1) 7 55
− (p + 1)2 + (p + 1)
2880 576 288
( ( ))
1 −3
+ (p − 23) 1 +
36 p
( ( ))
1 −1
+ (2p − 25) 1 +
96 p

( ( )) ( ( ))  −1/2 if p ≡ 1 mod 8
1 −1 −3
− 1+ 1+ + −1/4 if p ≡ 3, 5 mod 8
12 p p 
0 of p ≡ 7 mod 8

 −4/5 if p ≡ 1 mod 5
+ 0 if p ≡ 2, 3, 4 mod 5

−1/5 if p = 5.

p 2 3 5 7 11 13 17 19 23 29 31 37
dim 0 0 0 0 0 0 1 1 2 4 4 9

6
Theorem 2.3 For the Iwahori subgroups for B(p), we have dim S3 (B(2)) =
dim S3 (B(3)) = 0, and for p ≥ 5, we have

dim S3 (B(p)) =

(p + 1)2 (p2 + 1) 13 1
− (p + 1)2 + (p + 1)
2880 288 3
( ( )) ( ( ))
5p − 37 −1 3p − 29 −3
+ 1+ + 1+
48 p 6 p
( ( )) ( ( ))
1 −1 −3
− 1+ 1+
6 3 p
{
−1 if p ≡ 1 mod 8
+
0 otherwise

 −8/5 if p ≡ 1 mod 5
+ 1/5 if p = 5

0 otherwise

p 2 3 5 7 11 13 17 19 23 29 31 37
dim 0 0 0 0 2 9 25 42 88 237 312 649
′ ′
Theorem 2.4 For the Klingen type discrete subgroups Γ0 (p), we have dim S3 (Γ0 (2)) =
′
dim S3 (Γ0 (3)) = 0, and for p ≥ 5, we have
′
dim S3 (Γ0 (p)) =

(p + 1)(p2 + 1) 1 43
− (p + 1)2 + (p + 1)
2880 96 288
( ( )) ( ( ))
p − 11 −1 p − 13 −3
+ 1+ + 1+
32 4 18 p
{
−1/2 if p ≡ 1 mod 8
+
0 otherwise

 −4/5 if p ≡ 1 mod 5
+ 1/5 if p = 5

0 otherwise

7
 p 2 3 5 7 11 13 17 19 23 29 31 37
dim 0 0 0 0 0 1 1 2 2 5 6 13
We define

dim Sknew (B(p)) =

dim Sk (B(p)) − dim Sk (Γ0 (p))
′
−2 dim Sk (Γ0 (p)) + 2 dim Sk (Sp(2, Z))
+ dim Sk (K(p))

Then it is known that dim S3new (B(p)) is the dimension of the space of
F ∈ S3 (B(p)) whose corresponding local representation at p is the Stein-
berg representation.

Theorem 2.5 We have dim S3new (B(2)) = dim S3new (B(3)) = 0 and for
p ≥ 5, we have

dim S3new (B(p)) =

1 7
1+ (p − 1)(p3 − 1) − (p − 1)2
2880 ( ( )) 576 ( ( ))
1 −1 1 −3
− (p − 1) 1 − − (p − 1) 1 −
32 p 24 p
( ( )) ( ( ))
1 −1 −3
− 1− 1−
12 p p
( ( ))2
1 −3
− 1−
9 p


 1/5 if p = 5

2/5 if p ≡ 2, 3 mod 5
−

 4/5 if p ≡ 4 mod 5

0 otherwise
{
1/2 if p ≡ 7 mod 8
−
0 otherwise
{
1/6 if p ≡ 11 mod 12
−
0 otherwise

Numerical examples.

p 2 3 5 7 11 13 17 19 23 29 31
dim 0 0 0 0 2 8 23 38 83 225 298

8
2.4 Dimension formulas of weight 4
By combining the method (1) and (2) in Section 2, we can easily see that

Theorem 2.6 The dimension formulas of Sk (Γ) in [7], [14], [9] are valid
also for k = 4.

The more precise proof will be explained in the last section. For the
readers’ convenience, we write explicit results for k = 4 below. We obtain
the first one from [7] and the second one from [14] by putting k = 4 in the
formula there.
For Γ0 (p), we have dim S4 (Γ0 (2)) = 0, dim S4 (Γ0 (3)) = dim S4 (Γ0 (5)) = 1
and for any prime p > 5, we have
1 2 7 11
dim S4 (Γ0 (p)) = (p + 1)(p + 1) + (p + 1)2 − (p + 12)
576 ( ) 192 ( ) 288 ( )
1 −3 2p − 41 −1 1 3
+ (p − 1) + −
36 p 96 p 12 p
 
 4 if p ≡ 1 mod 8  4 if p ≡ 1 mod 12
1 1
+ × 2 if p ≡ 3, 5 mod 8 + × 2 if p ≡ 3, 5 mod 12
8  12 
0 if p ≡ 7 mod 8 0 if p ≡ 7 mod 12,
( )
−d
where is the Legendre symbol.
p
For K(p), we have S4 (K(2)) = S4 (K(3)) = 0 and for any prime p ≥ 5 we
have
( ) ( )
1 2 p−2 1 −1 p −3
dim S4 (K(p)) = (p + 1) + + (p − 12) +
576 8 96 p 36 p
( ) ( )
1 2 1 3
+ +
8 p 12 p
′ ′
We have dim S4 (Γ0 (2)) = dim S4 (Γ0 (3)) = 0. For any prime p ≥ 5, we
have
′ 1 2 5 23
dim S4 (Γ0 (p)) = (p + 1)(p + 1) − (p + 1)2 + (p + 1)
576 ( ( 288)) 288 ( ( ))
1 −1 1 −3
+ (7p − 41) 1 + + (3p − 7) 1 +
96 p 36 p
{ {
1 1 if p ≡ 1 mod 8 1 1 if p ≡ 1 mod 12
+ × + ×
2 0 otherwise 3 0 otherwise

9
 We have dim S4 (B(2)) = 0 and dim S4 (B(3)) = 1. For p ≥ 5, we have

dim S4 (B(p))

1 11 1
= (p + 1)2 (p2 + 1) + (p + 1)2 − (p + 1) + 1
576 ( ( 288)) 3( ( ))
3 −1 2 −3
+ (p − 5) 1 + + (p − 1) 1 +
16 p 9 p
( ( )) ( ( ))
1 −1 −3
− 1+ 1+
6 p p
{ {
1 if p ≡ 1 mod 8 2 1 if p ≡ 1 mod 12
+ + ×
0 otherwise 3 0 otherwise

We also have dim S4new (B(2)) = dim S4new (B(3)) = 0. For p ≥ 5, we have

dim S4new (B(p)) =

( ( ))
1 7 1 −1
(p − 1)(p − 1) +
3
(p − 1) − (p − 1) 1 −
2
576 192 32 p
( ( )) ( ( )) ( ( ))
1 −3 1 −1 −3
− (p − 1) 1 − − 1− 1−
18 18 12 p p
{ {
1 1 if p ≡ 7 mod 8 1 1 if p ≡ 11 mod 12
+ × + ×
2 0 otherwise 3 0 otherwise

Numerical examples of dimensions of weight 4 is given below.

p 2 3 5 7 11 13 17 19 23 29 31 37 41 43
S4 (K(p)) 0 0 0 1 1 2 2 3 3 4 6 8 7 9
S4 (Γ0 (p)) 0 1 1 3 7 11 20 27 41 75 90 143 185 211
′
S4 (Γ0 (p)) 0 0 0 1 1 5 7 10 14 34 43 79 101 118
S4 (B(p)) 0 1 2 9 33 70 176 269 545 1350 1753 3506 5220 6297
S4new (B(p)) 0 0 1 5 25 51 144 225 479 1211 1583 3213 4840 5859

3 Geometric Meanings of the Dimensions

An abelian surface A over an algebraically closed field k of characteristic p is
said to be super-singular if A is isogenous to E 2 where E is a super-singular
elliptic curve over k. We denote by A2,1 the moduli space of principally
polarized abelian surfaces.

10
Theorem 3.1 The number of irreducible components of the locus S of prin-
cipally polarized super-singular abelian surfaces in the moduli space A2,1 is
equal to
dim S3 (K(p)) + 1.

Theorem 3.2 The arithmetic genus of the locus S is equal to dim S3new (B(p)).

Remark. Each irreducible component of S is birational to P1 but has

many singularities. The generalized arithmetic genus of a singular reducible
curve was defined by J. P. Serre [24].
We obtain these theorems by applying the relation between arithmetic of
quaternion hermitian lattices and geometry, due to Katsura-Oort and partly
myself, and relation of dimension of Siegel modular forms and ”class num-
bers” of quaternion hermitian lattices obtained in [8]. In next section, we
explain this.

4 Arithmetic of Quaternion Hermitian Groups

Let D be the definite quaternion algebra over Q ramified exactly at p and ∞
and O a maximal order of D. We put

G = {g ∈ M2 (D); g t g = n(g)12 , n(g) > 0}.

Let GA be the adelization of G. The local factor at a place v ≤ ∞ is defined

by
Gv = {g ∈ M2 (Dv ) : g t g = n(g)12 , n(g) ∈ Q×
v }.

Here we understand that Q∞ = R and in that case we have n(g) > 0 au-
tomatically since D∞ is definite. In particular, G∞ /center is a compact
group isomorphic to {g ∈ M2 (H); g t g = 12 }/{±12 } where H is the Hamilton
quaternion algebra. For any prime q ̸= p, we have Gq ∼ = GSp(2, Qq ) = {g ∈
×
M4 (Qq ); gJ g = n(g)J}. We put Uq = GSp(2, Zq ) = GSp(2, Q
t
∏q ) ∩ M4 (Zq ) .
To define automorphic forms, we take a subgroup U = G∞ Up q̸=p Uq of GA ,
where we define a open compact group Up of Gp later. The open subgroup
U ⊂ GA plays the role of a ”discrete subgroup”. We take a representation
ρ : GA → G∞ → G∞ /center → GL(V ). An automorphic form with re-
spect to U of weight ρ is defined as a V -valued function of GA such that
f (uga) = ρ(u)f (g) for any u ∈ U , g ∈ GA and a ∈ G, where G is diagonally
embedded in GA as usual. We denote the space of these functions by Mρ (U ).
When ρ is the trivial representation, we write Mρ (U ) = M0 (U ). In this case,
the above definition implies that dim M0 (U ) = #(U \GA /G), which is called
the class number of U .

11
 Now by changing the coordinate a little, we take another group G∗p iso-
morphic to Gp , and under this identification we take Up as a subgroup Up,0 ,
Up,1 or Up,2 of G∗p as follows.

{ ( ) ( )}
0 1 t 0 1
Gp ∼
= G∗p = g ∈ M2 (Dp ); g g=
1 0 1 0
Up,1 = G∗p ∩ M2 (Op )×
( )×
∗ Op π −1 Op
Up,2 = Gp ∩
πOp Op
Up,0 = Up,1 ∩ Up,2 .

The group Up,0 is the minimal parahoric subgroup of G∗p . We denote by Ui

(i = 0, 1, 2) the corresponding open sugroup U of GA defined as before by
taking Up = Up,i . Then we have the following theorems.

Theorem 4.1 (Ib.-Katsura-Oort[10]) The number of principal polariza-

tions on E 2 is the class number of U1 .

Theorem 4.2 (Katsura-Oort [11]) The number of irreducible components

of supersingular locus S in A2,1 is the class number of U2 .

The following conjecture was suggested in late 1980’s in the discussion

with Professor Ekedahl.

Conjecture 4.3 The arithmetic genus of S is

dim M0 (U0 ) − dim M0 (U1 ) − dim M0 (U2 ) + 1.

In this occasion, we would like to add several more remarks on known

results about the arithmetic of quaternions and geometry.

(1) The configuration of S is described by inclusions of double cosets of

U0 gG, U1 hG, U2 kG.

(2) For each class U2 gi GA (1 ≤ i ≤ H), consider the finite groups Γi =

G∩gi−1 U gi . Each Γi is the automorphism group of the ”Moret-Bailey” family
(over the irreducible component of S). We can determine all Γi by a kind of
new mass formula. (cf. [15]).

(3) The curves C of genus 2 with J(C) ∼= E 2 have models defined over
Fp2 . Each number of curves C of genus 3 such that J(C) ∼
= E 2 defined over

12
Fp or defined over Fp2 is counted. (cf. [12]).

(4) For p ≥ 3, there exists a genus 3 curve C such that #(C(Fp2 )) attains
Weil’s maximum. (cf. [16]). It is an open problem if we can take the above
C as a hyper-elliptic curve, as far as the author knows.

Now, by the general philosoply of Langlands, or Ihara’s old conjecture

in [22], the relation of automorphic forms of GA should be related with au-
tomorphic forms of Sp(2, QA ). But here the concrete description of corre-
sponging discrete groups or new forms were not clear. The author gave more
concrete conjectures of this type correspondence in early 1980’s, aiming to
generalize classical theorems of Eichler betwee SL(2) and SU (2). We shortly
review them now. We denote by ρf1 ,f2 the finite dimensional representation
of Sp(2) = {g ∈ M2 (H); g t g = 12 } corresponding to the Young diagram
parameter f1 ≥ f2 ≥ 0 with f1 ≡ f2 mod 2.

Conjecture 4.4 ([14]) For any even non-negative integer j and any integer
k ≥ 3, there should exist an isomorphism

Mρnew
k+j−3,k−3
(U2 ) ∼ new
= Sk,j (K(p))

which preserves L functions.

Conjecture 4.5 ([13]) There should exist an isomorphism

Mρnew
k+j−3,k−3
(U0 ) ∼ new
= Sk,j (B(p)).

Here we do not explain the meaning of new forms in detail in general case
(see the above quoted references.) But when k = 3 and j = 0, then we define
M0new (U ) is space of automorphic forms orthogonal to the constant functions
and S3new (K(p)) = S3 (K(p)) (i.e. there are no old forms in this case).

5 Outline of the Proofs

We use the Riemann-Roch-Hirzebruch theorem and the holomorphic Lef-
schetz theorem. For a discrete subgroup Γ in question, we take a torsion
′
free normal subgroup Γ of Γ with finite index. We denote by X the Sa-
′
e a smooth toroidal compactification of
take compactification of Γ \H2 and X
X. We put D = X e − X. Then D is a divisor with simple normal cross-
ing. Let L be a holomorphic line bundle which is a natural prolongation of
′ ′
Γ \H2 × C where Γ acts on H2 × C by (Z, u) → (M Z, det(CZ + D)u). Let

13
 e We have L3 = Ω3 ⊗ [D] and
Ω be a sheaf of holomorphic 1 forms on X.
−1 ′
Ω = L ⊗ [D] . We put G = Γ/Γ and for any right G-module M , we put
3 3

e Ω3 ) = S3 (Γ′ ) and hence we have

M G = {m ∈ M ; m = mg}. We have H 0 (X,
e Ω3 )G = S3 (Γ).
H 0 (X,
e Ω3 )G . For any holomorphic vector bundle V on
So we must calculate H 0 (X,
e we put
X,
∑3
e
χ(X, V ) = e V ).
(−1)i dim H i (X,
i=0
This is called the Euler-Poincare characteristic of V . The formula to give this
number is called the Riemann-Roch-Hirzebruch theorem. The G-invariant
version of this theorem is given by the holomorphic Lefschetz theorem. This
is a formula to calculate
∑
3
e V)=
χG (X, e V )G .
(−1)i dim H i (X,
i=0

We have dim H i (X, e Ω3 × Lk−3 )G = dim Sk (Γ). To calculate dim H 0 (X, e V)

for V = Ω ⊗ L , we must do two things.
3 k−3

(1) Calculation of this altenating sum χG (X,e V ).

e V )G for i =
(2) Calculation of dim H i (X, ̸ 0.
There are done roughly as follows. We know already the formula for (1) for
big k by Selberg trace formula, and we can use this value also for smaller k as
e Ω3 )G = H 2 (X,
we shall see later. As for (2), we shall show that H 1 (X, e Ω3 ) =
e Ω3 )G ∼
0 for our discrete subgroups in question. It is also easy to see H 3 (X, =
e
C. So as a whole we have dim S3 (Γ) = χG (X, Ω ) + 1.
3

We shall see these more in detail. First we explain the calculation (1).
We review the holomorphic Lefschetz theorem (cf. [2]). This is given by the
following formula.
∑
3
1 ∑
e V )G =
(−1)p dim H p (X, τ (g)
i=0
|G| g∈G
∑
τ (g) = τ (g, Xαg )
{α ∏ θ g }
ch(V |X g
α ) · g U (Nα (θ)) · T (X g
α )
τ (g, Xαg ) = [Xαg ]
det(1 − g|(Nαg )∗ )

Here, for g ∈ G, X g ⊂ X denotes the fixed point set of g. Let Xαg be an

irreducible component of X g . We denote by Nαg the normal bundle of Xαg

14
and by (Nαg )∗ the dual bundle. We write
∑
V |Xαg = ai ⊗ χi ai ∈ K(Xαg ), χi ∈ R(G),
i

where K(Xαg ) is the Grothendieck group of ventor bundles and R(G) is the
representation ring over∑ C. We denote by ch(∗) the Chern character of ∗
and ch(V |Xα )(g) = i χi (g)ch(ai ). Now g acts also on the normal bundle
g

Nαg . Since g is of finite order, the eigenvalues of g of this action are roots of
unity and we denote them by eiθ . We decompose Nαg into eigenspaces and
write the corresponding bundle by Nαg (θ). We decompose the total Chern
class formally as follows.

c(Nαg (θ)) = 1 + c1 (Nαg (θ)) + · · · + cn (Nαg (θ))

∏
= (1 + xβ ).
β

Then we define
∏ ( 1 − e−xβ −iθ )−1
U θ
(Nαg (θ)) = .
β
1 − e−iθ

We denote by T (Xαg ) the Todd class of Xαg and by [Xαg ] the fundamental
class in H2d (Xαg , Z) where d = dim Xαg .
When V is a line bundle as in our case, then there is no decomposition
of the representation and we have just a character of g. If we take L⊗k , then
the action is the k-th power of the action on L. So if we put L|Xαg = a ⊗ χ,
then we have L⊗k |Xαg = a⊗χki where a is a line bundle on Xαg . So in this case
we have ch(L⊗k |Xαg ) = ch(a)k χk (g). Since g is of finite order, χ(g) are roots
of unity. In the holomorphic Lefschetz formula for V = Lk − D, the only
part which depends on k is the term ch(V |Xαg )(g). By the usual Rieman-
Roch-Hirzebruch theorem, this part is expressed with the product of χk (g)
and polynomials of k with coefficients which are independent of k. So we can
conclude that

Lemma 5.1 There exists a certain natural number M such that for each
e Ω ⊗ Lk−3 ) is a polynomial of k with constant coeffi-
fixed k mod M , χG (X,
cients.

Corollary 5.2 If we have a formula for Sk (Γ) for all k >> 4, then the same
formula gives the G-invariant Euler-Poincare characteristic for any k ≥ 3.

In other words, for any weight k, the holomorphic Lefschetz formula gives
e Ω⊗
us a formula of the ”G-invariant Euler-Poincaré characteristics” χG (X,

15
Lk−3 ) as polynomials of k for each k mod M for some big M . On the other
hand, it is known that χG (X, e Ω ⊗ Lk−3 ) = dim Sk (Γ) for any k ≥ 4. (cf.
Tsushima [27]). For discrete groups Γ of parahoric type, dim Sk (Γ) has been
calculated by Selberg trace formula in [7], [14], [9] (except for the case Γ =
′
B(3) or Γ0 (3), the case excluded in [9] by complication of calculation. But
′
we can show S3 (B(3)) = S3 (Γ0 (3)) = 0 by other ad hoc argument.) So the
first calculation is done.
Next we explain (2) for V = Ω3 . By Serre duality, we have
e Ω3 )G ∼
H i (X, e O)G .
= H 3−i (X,
By this, we have
e Ω3 )G ∼
H 3 (X, e O)G = C.
= H 0 (X,
By Dolbeault-Hodge theorem, we have
e O)G ∼
H 3−i (X, e Ω3−i )G .
= H 0 (X,
The left hand side is (H 0,3−i )G and the right hand side (H 3−i,0 )G is the
complex conjugation. Hence we have
e Ω3 )G ∼
H 1 (X, = H 0 (X, e Ω2 (logD)) ∼
e Ω2 )G ⊂ H 0 (X, = Adet ·Sym2 (Γ)

e Ω)G ∼
H 2 (X, = H 0 (X, e Ω(logD)) ∼
e Ω1 )G ⊂ H 0 (X, = ASym2 (Γ)
e Ω1 ) = 0 first by Takayuki
The author was informed of the fact H 0 (X,
Oda. He explained the author that we can show this by standard theorems
on cohomology and rather folklore. Here we can give an alternative proof
using the following theorem by Freitag.
Theorem 5.3 (Freitag [3]) For any congruence group Γ of Sp(2, Z), we
have ASym2 (Γ) = 0.
Freitag’s proof uses various modular embeddings of Hilbert modular forms
and induce a contradiction. Anyway, we have H 2 (X, e Ω3 ) = 0.
We will explain that Adet ·Sym(2) (Γ) = 0 for our Γ in the next section. If
we admit this, we can conclude as follows.
′ ′′
Theorem 5.4 For a discrete group Γ = K(N ), Γ0 (N ), Γ0 (N ), Γ0 (N ), B(p),
where N is any squarefree natural number and p is any prime, we have
e Ω3 ) + 1.
dim S3 (Γ) = χG (X,
As we explained, we get explicit value of χG (X, e Ω3 ) by putting k = 3 in
the general formula for big k in [7], [14], [9]. So we have dimension formulas
of weight 3.

16
6 Vanishing of Weight One
Here we sketch the proof of the following theorem.

Theorem 6.1 For any j, we have AdetSymj (Γ) = 0 for Sp(2, Z), K(N ),
′
Γ0 (N ), Γ0 (N ), B(p), as far as N is a squarefree natural number and p is a
prime.

For Γ = Γ0 (N ) and j = 0, this theorem is already in the joint work with

Skoruppa(cf. [20]) (though there was an error for the vanishing of Jacobi
forms and the condition that N is (at least) squarefree should be added.
Actually Lemma claim holds for a little more general N . ) The results for
j > 0 or other groups are new. We do not know in general if S1,j (B(N )) = 0
for natural numbers N which are not primes.
Proof is essentially based on the following result on Jacobi forms.
(1)
Theorem 6.2 ([20]) We have J1,m (Γ0 (N )) = 0 if N is squarefree and m
is coprime to N .

Note that in [20] contains an error. There it was claimed that this holds
for any N , but we need some conditions on N including all squarefree cases.
(1)
Here Γ0 (N ) is the usual subgroup of SL2 (Z). In particular, if N = 1 i.e.
for Γ0 (1) = SL2 (Z), we have J1,m = 0 always. This was known already in
Skoruppa [25].
So we review Jacobi forms here shortly. Let Γ1 be a finite index subgroup
of SL2 (Z). A holomorphic function f (τ, z) of H1 × C is said to be a Jacobi
form of weight k of index m belonging to ΓJ1 = Γ·Z2 if it satisfies the following
conditions (1) (2) (3).
(1) We have
( ) ( )
aτ + b z k m cz 2
f , = (cτ + d) e f (τ, z)
cτ + d cτ + d cτ + d
( )
a b
for any ∈ Γ1 , where we write em (x) = e2πimx .
c d
(2) We have
f (τ, z + λτ + µ) = em (−τ λ2 − 2λz)f (τ, z)
for any λ, µ ∈ Z. ( )
a b
(3) For any M = ∈ SL2 (R), we put
c d
( ) ( )
−k mcz 2 aτ + b z
f |k,m M = (cτ + d) e − f , .
cτ + d cτ + d cτ + d

17
Then for any M ∈ SL2 (Z), we have the following Fourier expansion.
∑
(f |k,j M )(τ, z) = c(n, r)e(nτ )e(rz)
n,r∈Z

where c(n, r) = 0 unless 4nm − r2 ≥ 0.

The space of such Jacobi forms are denoted by Jk,m (Γ0 (N )J ).
Now put  
Q 0 Q Q
Q Q Q Q 
P1 =  
Q 0 Q Q ∩ Sp(2, Q).
0 0 0 Q
Let Γ be a subgroup of Sp(2, Q) such that Γ ∩ P1 contains all
 
1 0 0 µ
λ 1 µ κ 
 
 0 0 1 −λ
0 0 0 1
with λ, µ, κ ∈ Z and  
a 0 b 0
0 1 0 0
 
c 0 d 0
0 0 0 1
( )
a b
with ∈ Γ1 . For any F ∈ Sk,j (Γ), we write the Fourier-Jacobi expan-
c d
sion as
∑
∞
F (Z) = fm (τ, z)em (ω)
m=1
( )
τ z
for Z = ∈ H2 . Here fm (τ, z) is a Cj+1 -valued function on H1 × C.
z ω
We denote by ϕm (τ, z) is the last component of fm (τ, z). We can show that
ϕm (τ, z) is a Jacobi form of weight k of index m belonging to ΓJ1 in the usual
sense. This can been seen by the relation
( ) ( )
aτ + d z k m cz 2 cτ + d cz
fm , = (cτ + d) e ( )ρj fm (τ, z)
cτ + d cτ + d cτ + d 0 1
( )
1 −λ
fm (τ, z + λτ + µ) = e (−λ τ − 2λz)
m 2
fm (τ, z)
0 1
and Koecher principle of F . In general, the other comonents are not Jacobi
forms, but obtained by derivatives of Jacobi forms (cf. [21]).

18
 So, for Γ = Γ0 (N ) or B(N ), the last component ϕm (τ, z) of the m-th
(1) ′
Fourier Jacobi coefficients belongs to Jk,m (Γ0 (N )J ). For Γ = Γ0 (N ), we
have ϕm ∈ Jk,m (SL2 (Z))J . When Γ = K(N ), since κ ∈ N −1 Z, ϕm is not
zero only when N |m. So if we renumber the coefficients as ϕm/N , then ϕm ∈
Jk,N m (SL2 (Z)J ).
The following lemma is obtained easily by seeing the automorphic prop-
erty of Siegel modular forms.

Lemma 6.3 Let Γ be one of the above discrete groups. If the last component
of F ∈ Sk,j (Γ) is identically zero, then F itself is identically zero.

Since we have J1,m (SL2 (Z)J ) = 0 for any m, this lemma implies imme-
′
diately that S1,j (Γ0 (N )) = S1.j (K(N )) = S1,j (Sp(2, Z)) = 0 for any j ≥ 0.
′ ′′ ′′
Since Γ0 (N ) and Γ0 (N ) are conjugate, we also have S1,j (Γ0 (N )) = 0. As for
the claim that S1,j (Γ0 (N )) = 0 for any natural number N and S1,j (B(p)) = 0
for any prime p with j ≥ 0, we need more argument similar to those as in
[20]. We omit the details of the proof here.

Correction: In the paper [9], there are following typos.

p.44 l.2; ”[−1, −k + 1, −k + 2, k − 1, k − 2; 6]” should read
[−1, −k + 1, −k + 2, 1, k − 1, k − 2; 6].
p.44 l.10; In the right hand side of t(β̂5 , k)+t(β̂6 , k), ”(−4/9)(1+(−1/p))[1, −1, 0; 3]”
should read (−4/9)(1 + (−3/p))[1, −1, 0; 3].
p.47 l.21; In the right hand side of T6 , ”[−k + 1, −k + 2; 2]” should read
[−k + 2, −k + 1; 2].
′
p.49 l.1; In the right hand side of H3 (U1 (p)), ”(p − 1)(1 − (−1/p))/26 /32 ”
should read (p − 1)(1 − (−1/p))/24 /3.
′ ′
p. 75; In (5.17), ”= m(γ; Sp(2, Z)/Γ0 (p))” should read m(γ; Sp(2, Z)/Γ0 (p)) =
′
p + 2 + (−1/p). Similarly, in (5.18), ”= m(γ; Sp(2, Z)/Γ0 (p))” should read
′
m(γ; Sp(2, Z)/Γ0 (p)) = p + 2 + (−3/p).

References
[1] A. Ash, D. Mumford, M. Rapoport and Y. Tai, Smooth compactifica-
tion of Local Symmetric Varieties, (Lie Groups: History, Frontiers, and
Application, Vol. 4). Math. Sci. Press, Brookline MA, 1975.

[2] M. F. Atiyah and I. M. Singer, The index of elliptic operators III, Ann.
of Math. (2) 87(1968), 546–604.

19
 [3] E. Freitag, Ein Verschwindungssatz für automorphe Formen zur
Siegelschen Modulgruppe. (German) Math. Z. 165 (1979), no. 1, 11–
18.

[4] E. Freitag, Siegelsche Modulfunktionen, Grundlehren der Mathematis-

chen Wissenschaften 254, Springer-Verlag, Berlin, 1983. x+341 pp.

[5] D. Mumford, Hirzebruch’s Proportionality Theorem in the non-compact

case, Invent. Math. 42(1977), 239–272.

[6] R. C. Gunning and H. Rossi, Analytic functions of several complex vari-

ables, Prentice Hall Inc. Englewood Cliffs, N. J. pp.317 +xii.

[7] K. Hashimoto, The dimension of the spaces of cusp forms on Siegel upper
half-plane of degree two. I. J. Fac. Sci. Univ. Tokyo Sect. IA Math. 30
(1983), no. 2, 403–488.

[8] K. Hashimoto and T. Ibukiyama, On class numbers of positive definite

binary quaternion Hermitian forms, (I) J. Fac. Sci. Univ. Tokyo Sect. IA
(1980), 549–601, (II) ibid., (1982), 695–699, (III) ibid., (1983), 393–401.

[9] K. Hashimoto and T. Ibukiyama, On relations of dimensions of auto-

morphic forms of Sp(2, R) and its compact twist Sp(2). II. Automor-
phic forms and number theory (Sendai, 1983), 31–102, Adv. Stud. Pure
Math., 7, North-Holland, Amsterdam, 1985.

[10] T. Katsura, T. Ibukiyama and F. Oort, Supersingular curves of genus

two and class numbers. Compositio Math. 57 (1986), no. 2, 127–152.

[11] T. Katsura and F. Oort, Families of supersingular abelian surfaces. Com-

positio Math. 62 (1987), no. 2, 107–167.

[12] T. Katsura and T. Ibukiyama, On the field of definition of superspe-

cial polarized abelian varieties and type numbers. Compositio Math. 91
(1994), no. 1, 37–46.

[13] T. Ibukiyama, On symplectic Euler factors of genus two. J. Fac. Sci.

Univ. Tokyo Sect. IA Math. 30 (1984), no. 3, 587–614

[14] T. Ibukiyama, On relations of dimensions of automorphic forms of

Sp(2, R) and its compact twist Sp(2). I. Automorphic forms and number
theory (Sendai, 1983), 7–30, Adv. Stud. Pure Math., 7, North-Holland,
Amsterdam, 1985

20
[15] T. Ibukiyama, On automorphism groups of positive definite binary
quaternion Hermitian lattices and new mass formula. Automorphic
forms and geometry of arithmetic varieties, 301–349, Adv. Stud. Pure
Math., 15, Academic Press, Boston, MA, 1989.

[16] T. Ibukiyama, On rational points of curves of genus 3 over finite fields.

Tohoku Math. J. (2) 45 (1993), no. 3, 311–329.

[17] T. Ibukiyama, Siegel Modular Forms of Weight Three and Conjectural

Correspondence of Shimura Type and Langlands Type, The conference
on L-functions, Fukuoka Japan 18-23 February 2006, edited by L. Weng
and M. Kaneko, World Scientific New Jersey London Singapore Beijing
Shanghai Hongkong Taipei Chennai (2007), 55-69,

[18] T. Ibukiyama, Paramodular forms and compact twist, Automorphic

Forms on GSp(4), Proceedings of the 9th Autumn Workshop on Num-
ber Theory, Ed. M. Furusawa, (2007), 37-48.

[19] T. Ibukiyama and H. Saito, On zeta functions associated to symmetric

matrices and an explicit conjecture on dimensions of Siegel modular
forms of general degree. Internat. Math. Res. Notices 1992, no. 8, 161–
169.

[20] T. Ibukiyama and N-.P.Skoruppa, A vanishing theorem of Siegel modu-

lar forms of weight one, to appear in Abhand. Math. Semi. Univ. Ham-
burg (2007).

[21] T. Ibukiyama and R. Kyomura, Vector valued Jacobi forms and vector
valued Siegel modular forms, Osaka J. Math. 48 (2011), 783-808.

[22] Y. Ihara, On certain arithmetical Dirichlet series. J. Math. Soc. Japan

16 1964 214–225.

[23] C. Poor and D. Yuen, Dimensions of cusp forms for Γ0 (p) in degree
two and small weights, Abhand. Math. Semi. Univ. Hamburg 77 (2007),
59–80.

[24] J. P. Serre, Groupes algébriques et corps de classes, Publications de

l’institut de mathématique de l’université de Nancago, VII. Hermann,
Paris 1959 202 pp.

[25] N-. P. Skoruppa, Über den Zusammenhang zwischen Jacobiformen

und Modulformen halbganzen Gewichts, Inaugural-Dissertation, Bon-
ner Mathematische Schriften 159, Bonn 1984

21
[26] R. Tsushima, A formula for the dimension of spaces of Siegel cusp forms
of degree three. Amer. J. Math. 102 (1980), no. 5, 937–977.

[27] R. Tsushima, An explicit dimension formula for the spaces of generalized

automorphic forms with respect to Sp(2, Z). Proc. Japan Acad. Ser. A
Math. Sci. 59 (1983), no. 4, 139–142.

[28] R. Tsushima, On dimension formula for Siegel modular forms. Auto-

morphic forms and geometry of arithmetic varieties, 41–64, Adv. Stud.
Pure Math., 15, Academic Press, Boston, MA, 1989.

Department of Mathematics
Graduate School of Science
Osaka University
Machikaneyama 1-16, Toyonaka
Osaka 560-0043 Japan
ibukiyam@math.wani.osaka-u.ac.jp

22