THETA SERIES, EISENSTEIN SERIES AND

POINCARÉ SERIES OVER FUNCTION FIELDS

Ambrus Pál
January 22, 2003.

Abstract. We construct analogues of theta series, Eisenstein series and Poincaré

series for function fields of one variable over finite fields, and prove their basic prop-
erties.

1. Introduction
Motivation 1.1. This paper is part of the project to prove analogues of the formu-
las of Gross-Zagier (see [8]) and Gross (see [7]) for function fields of one variable over
finite fields. Both papers [7] and [8] have two main parts. The first part consists
of the computation of certain pairings between Heegner points on modular curves,
so it is geometric in nature. The second part is analytical: approximately it is the
computation of Fourier coefficients of the unique holomorphic cuspidal newform of
weight 2 and level N whose Peterson product with any other holomorphic cuspidal
newform of weight 2 and the same level is equal to the central critical value (or
derivative) of a certain L-function of the latter.
The aim of this paper is to construct the analogues of the three special types
of automorphic forms appearing in these analytical computations and prove their
basic properties which we will need later. Partially because of its technical nature
we also tried to present this material in a simple and transparent way, therefore
we might spend more time with certain parts than it is strictly necessary using the
results available in the literature. The three special types of automorphic forms
mentioned above are theta series, Eisenstein series and Poincaré series. The first
two of these are necessary in the analogue of the Rankin-Selberg method. On the
other hand Poincaré series can be used to construct linear generators of different
linear sub-spaces of cusp forms, and so they can be used to compute the projection
of automorphic forms into these sub-spaces. Our results are extensions of the results
of [15], [16], and [17], where the special case of the rational function field is treated.
In [18] these results were used to prove an analogue the Gross-Zagier formula in this
special case with several other restrictions. Our methods are also similar, but it is
necessary to treat all objects form the adélic point of view for the generalization.
Contents 1.2. In chapter 2 we collect some basic definitions and results on au-
tomorphic forms and their Fourier coefficients together from [11] and [22] for the
convenience of the reader and to fix notation. The only new result is Proposition 2.8
which says that if the Fourier coefficients of a holomorphic family of automorphic
forms can be continued analytically then the family also has analytical continuation.
Typeset by AMS-TEX
1
2 AMBRUS PÁL

The aim of chapter 3 is to construct theta series and determine their conductor
and central character. There are several constructions of theta series in the liter-
ature: in [5] it is constructed using the Weil representation, in [15] the analogue
of the classical treatment is worked out in the special case of the rational function
field and unramified central character. In both cases it is somewhat cumbersome
to to show that the series, as defined, is invariant under the congruence group of
the required conductor. For example in [15] this is proved via a fairly complicated
computation using the Poisson summation formula. We will choose a third way
which is perhaps the simplest one: we apply Weil’s converse theorem to induced
representations to show the existence of attached automorphic forms. The determi-
nation of their conductor and central character follows easily from standard results
of representation theory of finite groups.
Chapter 4 contains some preliminary results, some with proofs, mainly for the
convenience of the reader. In chapter 5 we compute the Fourier coefficients of
certain Eisenstein series. Eisenstein series were studied in various generality in the
papers [9], [10] and [14]. In particular the authors computed the constant terms in
the Fourier expansion of these series, remarked that they are Hecke eigenforms, and
gave a formula without proof for the eigenvalues (at least for primes not dividing the
conductor) which could be used to reduce the problem above to compute just one
Fourier coefficient. Because it is neither more complicated nor longer to compute
all Fourier coefficients at once rather than to compute just one, and because we
do not wish to burden our paper with references, we decided to prove our result
independently of theirs.
The aim of chapters 6 and 7 is to study analogues of Poincaré series for function
fields. In chapter 6 after a preliminary result (Lemma 6.3) for which we could
not find an adequate reference, we prove the convergence of Poincaré series and
give a characterization using the Peterson product. In chapter 7 we prove our
main results which are two estimates of the Fourier coefficients of Poincaré series.
The estimates are reduced to an estimate of local integrals which are reduced to
estimates of Kloosterman sums. Again we gave a proof for the latter (Lemma 7.2)
because of the lack of adequate reference. Here the adélic approach is essential,
otherwise the computations would be too complicated.

Notation 1.3. In this paper, if not otherwise stated, we will use the following con-
ventions, assumptions and notation. For any geometrically connected non-singular
projective curve C over the finite field f = Fq of characteristic p let |C|, Div(C)
and Pic(C) denote the set of closed points, the divisor group and Picard group of
an algebraic curve C, respectively. We will identify the set |C| and the set of places
of the function field of C. For any x ∈ |C| let deg(x) denote its degree. For any
D ∈ Div(C) let [D] denote its linear equivalence class.
X will be a fixed geometrically connected smooth projective curve over the field
Fq . We denote the function field and the ring of adéles of X by F and A, respec-
tively. We denote by | · | the normalized absolute value of any adéle in A. We
denote by O the maximal compact sub-ring of A. For any closed point v in |X|
we let Fv , fv and Ov denote the corresponding completion of F , its constant field,
and its discrete valuation ring, respectively. Let (a) denote the divisor of any idéle
a ∈ A∗ . Sometimes we will drop the parentheses for the sake of convenience.
For any idéle, adéle, adéle-valued matrix or function defined on the above which
decomposes as an infinite product of functions defined on the individual components
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 3

the subscript v will denote the v-th component. For example | · |v will denote the
normalized absolute value of the field Fv . Similar convention will be applied to
subsets of adéles and adéle-valued matrices. Sometimes, when it would be too
inconvenient otherwise, we will drop the subscript v.
Let G denote the group scheme of invertible two by two matrices. Let B denote
the group scheme of invertible upper triangular two by two matrices. Let P denote
the group scheme of invertible upper triangular two by two matrices with 1 on the
lower left corner. Let U denote the group scheme of invertible upper triangular two
by two matrices with ones on the diagonal. Let Z denote the center of G.
Acknowledgment 1.4. I wish to thank my thesis adviser, Dorian Goldfeld and
Henri Darmon for their help and encouragement. This work is based on the author’s
thesis at Columbia University. It was written mainly while he was visiting the
Institute for Advanced Study at Princeton, and completed while he was a post-
doctoral fellow at CICMA. I also wish to thank all these academic institutions for
for their warm hospitality and the pleasant environment they created for productive
research. Finally I wish to thank the referee for his useful remarks. The author
was partially supported by NSF grant DMS 97-29992.

2. Automorphic Forms
Definition 2.1. By an automorphic form over F of central character ψ, (where ψ
is a quasi-character of A∗ ), we mean a continuous function φ : G(A) → C satisfying
the following properties:
(i) φ(γg) = φ(g) for all γ ∈ G(F ),
(ii) φ(gz) = φ(g)ψ(z) for all z ∈ Z(A),
(iii) φ is right G(O)-finite, i.e. its right translates with respect to the elements of
the maximal compact subgroup G(O) of G(A) generate a finite dimensional vector
space over C.
Moreover, if for almost all g ∈ G(A):
Z  
1 x
φ( g)dx = 0,
F \A 0 1
where dx is a Haar measure on A, we call φ a cusp form.
Definition 2.2. Let ψ be a quasi-character of A∗ whose conductor divides n, where
n is an effective divisor in Div(X). By an automorphic form over F of level n, and
central character ψ, we mean an automorphic form satisfying the property (iii)∗
instead of the weaker property (iii) above.
(iii)∗ φ(gk) = φ(g)ψ(k) for all k ∈ K0 (n), where
 
a b
K0 (n) = { ∈ G(O)|c ≡ 0 mod n},
c d
and the (well-defined) representation ψ : K0 (n) → C∗ is given by:
   
a b Y a b
ψ( )= ψv (dv ) (∀ ∈ K0 (n)),
c d c d
v|n

if n is not the trivial divisor and 1, otherwise.

Let A(n, ψ) (resp. A0 (n, ψ)) denote the space of automorphic forms (resp. cus-
pidal automorphic forms) of level n, and central character ψ.
4 AMBRUS PÁL

Lemma 2.3. Every φ ∈ A(n, ψ) is uniquely determined by its restriction to P (A).

Proof. See [22], page 19. 
2.4. Let τ : F \A → C∗ be a non-trivial additive character. Then every character
of F \A is of the form x 7→ τ (ηx) for some η ∈ F .
Proposition 2.5. For every φ ∈ A(n, ψ) there are functions φ0 : Pic(X) → C and
φ∗ : Div(X) → C, the latter vanishing on non-effective divisors, such that
 
y x X
φ( ) = φ0 (y) + φ∗ (ηyd−1 )τ (ηx),
0 1 ∗
η∈F

for all y ∈ A∗ and x ∈ A, where d is an idéle such that D = dO, where D is the
O-module defined as
D = {x ∈ A|τ (xO) = 1}.

Proof. This is Proposition 1 of Chapter III in [22], page 21. For proof see pages
19-20 in [22]. 
Definition 2.6. The functions φ0 and φ∗ are called the Fourier coefficients of the
automorphic form φ with respect to the additive character τ .
Lemma 2.7. Assume that two functions φ0 : Pic(X) → C and φ∗ : Div(X) → C
are given, the latter vanishing on non-effective divisors. Then the sum
 
y x X
φ( ) = φ0 (y) + φ∗ (ηyd−1 )τ (ηx)
0 1 ∗
η∈F

has only finitely many non-zero terms over every compact set of P (A), so it defines
a continuous function on P (A).
Proof. This is Lemma 2 of Chapter V in [22], page 35. For proof see pages 35-36
in [22]. 
Proposition 2.8. Let U ⊂ V be two non-empty sub-domains of C. Let ψ : A∗ →
C, φ0s : Pic(X) → C and φ∗s : Div(X) → C be a quasi-character of A∗ and two
functions, respectively, the latter two are holomorphic in s for all s ∈ V , and the
last function vanishing on non-effective divisors such that
 
y x X
φs ( ) = φ0s (y) + φ∗s (ηyd)τ (ηx)
0 1 ∗
η∈F

is a restriction of an automorphic form in A(n, ψ) for all s ∈ U . Then φs is a

restriction of an automorphic form in A(n, ψ) for all s ∈ V , and for all g ∈ G(A)
the function Φs (g) is a holomorphic in s.
Proof. We Q first prove that G(A) = G(F )B(A)K for every compact, open sub-
group K = v∈|X| Kv . Take any element g of G(A). There is a finite set S of
Q such that if Kv is not G(Ov ) then s ∈ S. As the natural image of G(F )
places
in v∈S G(Fv ) is dense, there is a γ ∈ G(F ) such that the v-component of γ −1 g
is in Kv for all v ∈ S. But γ −1 g is in B(Fv )Kv = B(Fv )G(Ov ) for all other
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 5

v by the Iwasawa decomposition, so the claim follows. Let us start the proof of
the proposition. The function φs is well defined for all s by Lemma 2.7. By
what we have just proved it is sufficient to prove that for every γ1 , γ2 ∈ G(F ),
b1 , b2 ∈ P (A), k1 , k2 ∈ K0 (n) and z1 , z2 ∈ Z(A), if we have γ1 b1 k1 z1 = γ2 b2 k2 z2 ,
then φs (b1 )ψs (k1 )ψ(z1 ) = φs (b2 )ψs (k2 )ψ(z2 ) for all s ∈ V . But we know this to
hold for all s ∈ U , so it holds for all s ∈ V as well by unique continuation of
holomorphic functions. The last assertion of the proposition is obvious. 

3. Theta Series
Notation 3.1. For any field K let K denote its separable closure. For any Galois
extension F2 |F1 of global or local fields of positive characteristic let W (F2 |F1 )
denote the Weil group of the extension. For any global or local field K of positive
characteristic let W (K) simply denote W (K|K), the absolute Weil group of K. For
any n-dimensional complex representation ρ : W (K) → GLn (C) of the absolute
Weil group of the global or local field K of positive characteristic let det(ρ), cond(ρ)
denote the determinant of ρ, and its Artin conductor, respectively. If the the field
K is a global function field, let Lx (ρ, t) denote the Euler factor of the Grothendieck
L-series of ρ at x for each place x of K. If the representation above is 1-dimensional
and the field K is a global function field, then let ρ also denote the corresponding
grössencharacter of K by abuse of notation.
Definition 3.2. We will call the divisors m, n on a non-singular projective curve
C defined over a finite field relatively prime if their support is disjoint. We will call
a function f : Div(C) → C multiplicative if it vanishes on non-effective divisors,
and for every pair of relatively prime divisors n and m we have f (nm) = f (n)f (m).
Let ρ be an n-dimensional complex representation of the absolute Weil group of
the function field of C. Then let ρ also denote by abuse of notation the unique
multiplicative function ρ : Div(C) → C such that for each prime divisor p ∈ |C| we
have
X∞
ρ(pn )tdeg(p)n = Lp (ρ, t).
n=0

Theorem 3.3. Let ρ be a 2-dimensional complex representation of the absolute

Weil group of F . Then there is an automorphic form L(ρ) of level cond(ρ) and
central character det(ρ) such that for each effective divisor m ∈ Div(X) we have
L(ρ)∗ (m) = ρ(m).
If the representation ρ is irreducible, then L(ρ) can be chosen to be cuspidal.
Proof. This is the theorem of §76 of Chapter XI in [22], and the remark immedi-
ately following it, page 157. For proof see pages 145-157 in [22]. 
Notation 3.4. Let ∆ be a subgroup of finite index of a group Γ. If χ : ∆ →
Aut(V ) is a representation of ∆ on a complex vector space V then we will denote
the representation induced from χ to Γ by IndΓ|∆ (χ). If χ : Γ → Aut(V ) is a
representation of Γ on a complex vector space V then the restriction of χ to ∆
will be denoted by Res∆ (χ). For any element of g ∈ Γ the conjugate subgroup
g∆g −1 will be denoted by ∆g . For any representation χ : Γ → Aut(V ) as above we
will denote by χg the conjugate representation defined by χg (h) = χ(ghg −1 ) for all
h ∈ Γ.
6 AMBRUS PÁL

Lemma 3.5. If ∆ is a subgroup of a finite group Γ and χ : ∆ → Aut(V ) is a

finite dimensional representation on a complex vector space V , then the determinant
of the induced representation IndΓ|∆ (χ) is s(π)m t(det(χ)), where π is the regular
permutation representation of Γ on the left cosets of ∆, s(π) is its sign, t the map
induced on the dual groups by the transfer map Γab → ∆ab , and m = dim(V ).
Proof. See [2], Proposition 13.15, part (i), pages 338-339. 
Lemma 3.6. Let ∆, Ω be two subgroups of a finite group Γ and let χ : ∆ → Aut(V )
be a finite dimensional representation on a complex vector space V . Let S ⊂ Γ be
a representative system of double cosets of the pair ∆, Ω. Then
M
ResΩ (IndΓ|∆ (χ)) = IndΩ|Ω∩∆g−1 (ResΩ∩∆g−1 (χg )).
g∈S

Proof. See [2], Theorem 10.13, pages 237-238. 

Notation 3.7. Let Y be a separable connected double cover of X, and let K denote
the the corresponding extension of F . Let  : W (F ) → C∗ denote the unique non-
trivial character with kernel W (K) < W (F ). Let f denote the conductor of , the
discriminant of the extension K|F . Let N(m) ∈ Div(X) denote the norm of any
effective divisor m ∈ Div(Y ).
Proposition 3.8. Let λ : W (K) → C∗ be a continuous one-dimensional represen-
tation of the Weil group of K. Then the following holds:
(i) IndW (F )|W (K) (λ) is irreducible if and only if λ is not a restriction of a repre-
sentation of W (F ),
(ii) the determinant of IndW (F )|W (K) (λ) is t(deg(ρ)), where t denotes the same
homomorphism as in Lemma 3.5,
(iii) the Artin conductor of IndW (F )|W (K) (λ) is fN(cond(λ)),
(iv) for all x ∈ |X| we have
Y
Lx (IndW (F )|W (K) (λ), t) = Ly (λ, tdeg(y)/ deg(x) ),
y|x

where the product is over all y places of K above x.

Proof. (i) By Criterion 10.25 in [2], page 245, this induced representation is ir-
reducible if and only if λ is not equal to λg for any g ∈ W (K) − W (F ), since
two 1-dimensional representations are orthogonal if and only if they are different.
Clearly, the latter holds if and only if λ is not a restriction of a representation of
W (F ).
(ii) Let L be a normal extension of K such that the representation λ factors through
the finite quotient W (L|K). This group is a two index subgroup of W (L|F ). Since
the sign representation of the permutation of W (L|F ) on cosets of W (L|K) is ,
the claim follows from the Lemma 3.5 applied to λ.
(iii) This claim is a special case of Proposition 6 of chapter 6 in [19], page 111.
(iv) Let L denote the same normal extension as above. Let z be a place of L above
x. Let Γ, ∆ and Ω denote the Weil groups W (L|F ), W (L|K) and the decomposition
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 7

group W (Lz |Fx ), respectively. Then by Lemma 3.6 we have

M
ResΩ (IndΓ|∆ (λ)) = IndΩ|Ω∩∆g−1 (ResΩ∩∆g−1 (λg ))
g∈S
M −1
= IndΩg |Ωg ∩∆ (ResΩg ∩∆ (λ))g ,
g∈S

where S is a representative system of double cosets of the pair ∆, Ω. The group Γ

acts on all places of L above x transitively such that the action on the corresponding
decomposition groups is by conjugation. For every place y of K above x there is a
unique g ∈ S such that z g is above y. These facts together imply that
Y −1
Lx (IndW (F )|W (K) (λ), t) = Lx (IndΩg |Ωg ∩∆ (ResΩg ∩∆ (λ))g , t)
g∈S
Y
= Ly (λ, tdeg(y)/ deg(x) ),
y|x

where in the second equation we used (ii) of Proposition 3.8 on page 530 in [3]. 
Remark 3.9. A slight modification of the argument in §3 of Chapter XII in [13],
pages 236-239, can also be used to give a proof for part (iv) of the proposition
above.
Theorem 3.10. Let λ : W (K) → C∗ be a continuous one-dimensional repre-
sentation of the Weil group of K. Then there exists an automorphic form Θλ of
conductor N(cond(λ))f and central character t(λ) such that
X
Θ∗λ (n) = λ(m).
N(m)=n

If λ is not a restriction of a representation of W (F ), then Θλ can be chosen to be

cuspidal.
Proof. Let Θλ denote the automorphic form L(ρ) attached to the 2-dimensional
induced representation ρ = IndW (F )|W (K) (λ) in Theorem 3.3. It is clear that Θλ has
the required conductor and central character, and it can be chosen to be cuspidal,
by (iii), (ii) and (i) of Proposition 3.8, respectively. To conclude the proof of the
theorem, we only have to show that
X
IndW (F )|W (K) (λ)(n) = λ(m)
N(m)=n

for each effective divisor n of X. Because both functions in the equality above are
multiplicative, we only have to check this for prime divisors, but this case follows
at once from (iv) of Proposition 3.8. 

4. Some Preliminary Results

Definition 4.1. Let ν1 , ν2 be two continuous 1-dimensional complex representa-
tions of O∗ which are trivial on F∗q . Let π(ν1 , ν2 ) denote the right-regular repre-
sentation of G(O) on B(ν1 , ν2 ), where the latter is the set of all locally constant
complex-valued functions f on G(O) such that
   
a b a b
f( g) = ν1 (a)ν2 (c) · f (g)(∀ ∈ B(O), ∀g ∈ G(O)).
0 c 0 c
8 AMBRUS PÁL

Let ω be a grössencharacter of F such that composition of ω with the natural map

O∗ → F ∗ \A∗ is ν1 ν2 . Let Λ(ν1 , ν2 , ω) denote the set of all pairs (λ1 , λ2 ) of quasi-
characters of A∗ such that the composition of λ1 , λ2 with the natural map above
is ν1 , ν2 , respectively and λ1 λ2 = ω. As a principal homogeneous set over the
group of quasi-characters of Pic(X), the set Λ(ν1 , ν2 , ω) has a natural structure of
a Riemann surface. For any pair (λ1 , λ2 ) ∈ Λ(ν1 , ν2 , ω), let Re(λ1 λ−1 2 ) denote the
2Re(λ1 λ−1
real number such that for any idéle i ∈ A∗ we have |λ1 λ−1 2 (i)| = |i| 2 ).

Definition 4.2. Let ν1 , ν2 and ω be as above and let (λ1 , λ2 ) ∈ Λ(ν1 , ν2 , ω) be arbi-
trary. Let π(λ1 , λ2 ) denote the right-regular representation of G(A) on B(λ1 , λ2 ),
where the latter is the set of all locally constant complex-valued functions f on
G(A) such that
   
a b a b
f( g) = λ1 (a)λ2 (c) · f (g)(∀ ∈ B(A), ∀g ∈ G(A)).
0 c 0 c

Since by the Iwasawa decomposition the restriction map f 7→ f |G(O) induces

a G(O)-equivariant isomorphism between B(λ1 , λ2 ) and B(ν1 , ν2 ), we note that
π(λ1 , λ2 ) as a representation of G(O) is isomorphic to π(ν1 , ν2 ). For every φ ∈
B(ν1 , ν2 ) and (λ1 , λ2 ) ∈ Λ(ν1 , ν2 , ω), let φ(λ1 , λ2 ) denote the corresponding ele-
ment in B(λ1 , λ2 ).
Proposition 4.3. Let ν1 , ν2 and ω be as above.
(i) For all φ ∈ B(ν1 , ν2 ), the series
X
E(λ1 , λ2 )(φ)(g) = φ(λ1 , λ2 )(γg),
γ∈B(F )\G(F )

is locally uniformly absolute convergent on the open set Re(λ1 λ−1 2 ) > 1 of the
Riemann surface Λ(ν1 , ν2 , ω) and defines an automorphic form of central character
ω,
(ii) The series E(λ1 , λ2 )(φ)(g) is holomorphic in (λ1 , λ2 ) on the set above,
(iii) The map φ(λ1 , λ2 ) 7→ E(λ1 , λ2 )(φ) is G(A)-equivariant.
Proof. Claim (ii) and (iii) are obvious, as well as the second part of claim (i).
To prove the other part of (i), we first note that it is sufficient to prove it only
for E(| · |s , | · |−s )(χ), where χ is the function which is identically one on G(O).
To see this first note that for every φ ∈ B(ν1 , ν2 ) there is a constant M such that
|φ(k)| ≤ M for every k ∈ G(O) as that group is compact and φ is continuous. Then
we have the estimate
   
a b a b
|φ(λ1 , λ2 )( k)| ≤ M |λ1 (a)λ2 (c)| = M χ(| · |s , | · |−s )( k),
0 c 0 c
 
a b
where Re(λ1 λ−1
2 ) = s, for all ∈ B(A) and for all k ∈ G(O) using the
0 c
definition B(λ1 , λ2 ). But G(A) = B(A)G(O) by the Iwasawa decomposition so the
estimate holds for all g ∈ G(A). On the other hand the series E(| · |s , | · |−s )(χ) is
locally uniformly absolute convergent for Re(s) > 1 by Theorem 1.4.4. of [9], page
264. 
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 9

Definition 4.4. Let X be a Hausdorff topological space. A Borel measure on X

is a positive measure on the Borel sets of X. The latter will be denoted by B(X).
If Γ is a group which acts on X continuously on the left, then we say that µ is
invariant respect to Γ, if for every Borel set S ∈ B(X) and every γ ∈ Γ we have
µ(S) = µ(γS). Let ∆ be a subgroup of Γ. We say that the action of Γ on X relative
to ∆ is discrete, if the restriction of the action to ∆ is trivial and every x ∈ X has
an open neighborhood U such that if for some γ ∈ Γ we have that U ∩ γU 6= ∅,
then γ ∈ ∆.
Remarks 4.5. If a group Γ acts on a Hausdorff topological space X discreetly
and ∆ is a subgroup of Γ, then the set ∆\X inherits a natural Hausdorff topology
such that the projection map X → ∆\X is a covering, and Γ acts on ∆\X dis-
creetly relative to ∆. Since Z(F )\G(F ) acts on Z(A)\G(A) discreetly, it acts on
Z(A)U (F )\G(A) discreetly relative to Z(F )\U (F ).
Proposition 4.6. Let φ : Z(A)U (F )\G(A) be a continuous function which has
compact support. Then the series
X
P (φ)(g) = φ(γg),
γ∈Z(F )U (F )\G(F )

is uniformly absolute convergent and defines an automorphic form with trivial cen-
tral character which is compactly supported as a function on Z(A)G(F )\G(A).
Proof. Let S ⊂ Z(A)U (F )\G(A) be a compact subset containing the support of
φ. It is clear that if P (φ) is convergent, then its support is contained in the image
of S, hence it is compact. Therefore it is sufficient to prove that for each g ∈ G(A)
there is an open subgroup U < G(A) such that the number of γ ∈ Z(F )U (F )\G(F )
such that φ(γgu) is non-zero for some u ∈ U is finite. Since S is compact, it can
be covered by finitely many open subsets V such that if for some γ ∈ Z(F )\G(F )
we have that V ∩ γV 6= ∅, then γ ∈ Z(F )\U (F ). We can assume that all sets V
are right translates of an open subgroup U by refining the cover. This subgroup U
will suffice. 
The following proposition is an easy exercise in measure theory.
Proposition 4.7. Let X be a Hausdorff topological space which has a countable
basis of compact open subsets. Let µ be a Borel measure on X and let Γ be a group
which acts on X continuously on the right. Assume that µ is invariant respect to Γ.
Let ∆ be a subgroup of Γ such that the action of Γ on X relative to ∆ is discrete.
Then there is a unique measure on Γ\X, denoted by µ by abuse of notation
P the following property: if f : X → C is a continuous function such that
with
γ∈∆\Γ |f (γx)| is locally uniformly convergent and
Z X
|f (γx)|dµ < ∞,
Γ\X γ∈∆\Γ

then X
f (γx) ∈ L1 (Γ\X, µ) , f ∈ L1 (X, µ),
γ∈∆\Γ

and Z Z X
f (x)dµ = f (γx)dµ. 
X Γ\X γ∈∆\Γ
10 AMBRUS PÁL

5. Fourier Coefficients of Eisenstein Series

Definition 5.1. Let K denote a separable ramified extension of F of degree two,
and let , f denote the corresponding grössencharacter and its conductor, respec-
tively. Let Hs denote the unique complex-valued function on G(A) which satisfies
the following condition: for all g ∈ G(A) which is of the form
 
a b
g= k,
0 c

with  
a b
∈ B(A), k ∈ K0 (f),
0 c
we have
a
Hs (g) = | |s (c)(k),
c
and Hs (g) = 0, otherwise.
By Proposition 4.3 the series
X
Es (g) = Hs (γg),
γ∈B(F )\G(F )

is locally uniformly absolute convergent if Re(s) > 1, and defines an automorphic

form with central character  and level f.
Proposition 5.2. Let τ : F \A → C∗ be a non-trivial additive character such
that the divisor d defined in Proposition 2.3 has disjoint support from f. If Es0 and
Es∗ denote the Fourier coefficients of the automorphic form Es with respect to the
additive character τ , then for Re(s) > 1 we have

(yd)|d|3/2−s Y Y 1 − v (yv πv )|yv πv |2s−1

Es∗ (y) = κ v v (y v )|y v π v |2s−1
v
v
,
L(, 2s)|y|s−1
v|f
1 − v (πv )|πv |2s−1
v6|f
v

where κv is defined by
X
κv = |πv | v (a/πv )τv (−a/πv ),
a∈fv∗

πv is a local uniformizer for all v ∈ |X|, and Es0 (y) = |y|s .

Proof. By the definition of Fourier coefficients Es∗ (yd−1 ) is given by the formula
Z  
y x
Es∗ (yd−1 ) = |d|1/2 Es ( )τ (−x)dx,
F \A 0 1

where the product measure dx on A is chosen such that Ov has measure one for
all place v. As G(F ) has the pair-wise disjoint decomposition
  
[ [ 0 1 1 u
G(F ) = B(F ) B(F ) ,
1 0 0 1
u∈F
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 11

we get that
Z   Z  
y x y x
Es ( )τ (−x)dx = Hs ( )τ (−x)dx
F \A 0 1 F \A 0 1
Z    
0 1 y x
+ Hs ( )τ (−x)dx.
A 1 0 0 1

By definition the first term is equal to

Z
|y|s τ (−x)dx = 0.
F \A

Using the identity

     
0 1 y x 1 0 0 1
=
1 0 0 1 0 y 1 xy −1

and substituting x by xy −1 , the second integral can be written as

Z   
0 1 y x Y
Hs ( )τ (−x)dx = (y)|y|1−s (Vs )v (yv )
A 1 0 0 1
v∈|X|

where for y ∈ Fv we define

Z  
0 1
(Vs )v (y) = (Hs )v ( )τv (−xy)dx.
Fv 1 x

For any v ∈ |X| we have the formula


   v (x)|x|−2s
v if x ∈ / Ov ,
0 1
(Hs )v ( )= 1 if v 6 |f, x ∈ Ov ,
1 x
0 otherwise,


upon using the identity

x−1
    
0 1 1 −1 0
=
1 x 0 x x−1 1

in the first case.

From this it follows that
Z R
τ (−xyv )dx
Ov v
if v 6 |f,
(Vs )v (yv ) = v (x)|x|−2s
v τ v (−xyv )dx +
x∈O
/ v 0 otherwise.

We compute
Z XZ
v (x)|x|−2s
v τ v (−xy v )dx = v (xπv−n )|xπv−n |−2s
v τv (−xyv πv−n )dπv−n x
x∈O
/ v n≥1 Ov∗
X Z
= v (πv ) n
|πv |v(2s−1)n v (x)τ (−xyv πv−n )dx.
n≥1 Ov∗
12 AMBRUS PÁL

If v 6 |f, then since

Z 
1 if v(y) ≥ v(d),
τv (−xyv )dx =
Ov 0 otherwise,

and 
Z  1 − |πv |v if v(y) ≥ v(d) + n,
τv (−xyv πv−n )dx = −|πv |v if v(y) = v(d) + n − 1,
Ov∗ 
0 otherwise,
we obtain:
X
(Vs )v (y) =1 + v (πv )n |πv |v(2s−1)n (1 − |πv |v )
1≤n≤v(yd−1 )
−1
− |πv |v (v (πv )|πv |v(2s−1) )(v(yd )+1)

v(yd−1 )
X
=(1 − v (πv )|πv |2s
v ) v (πv )n |πv |v(2s−1)n
n=0
(2s−1) −1
1 − (v (πv )|πv |v )(v(yd )+1)
=(1 − v (πv )|πv |2s
v ) ,
1 − v (πv )|πv |2s−1
v

if v(y) ≥ v(d) and 0, otherwise.

If v|f, then

v (yv πv−n )κv

Z 
if v(y) = n − 1,
v (x)τv (−xyv πv−n )dx =
Ov∗ 0 otherwise.

From this we obtain:


v (yv )|yv πv |v2s−1 κv if v(y) ≥ 0,
Vs (y) =
0 otherwise.

Combining this together, we have (for y ∈ dO):

(y)|d|1/2 Y Y 1−v (yv d−1 πv )|yv d−1 πv |2s−1

Es∗ (yd−1 ) = κv v (yv )|yv πv |v2s−1 v v v
.
L(, 2s)|y|s−1
v|f
1 − v (πv )|πv |v2s−1
v6|f

The Fourier coefficient Es0 (y) is given by the formula

Z  
y x
Es0 (y) = |d| 1/2
|y| Es ( )dx.
F \A 0 1

Using the Bruhat-Tits decomposition of G(F ) again we can write this integral as
Z  
y x
Es0 (y)|d|−1/2 = Hs ( )dx
A/F 0 1
Z   
0 1 y x
+ Hs ( )dx.
A 1 0 0 1
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 13

The first term is equal to

Z
|y|s dx = |d|−1/2 |y|s .
F \A

Computing as above we can write the second term as

Z   Z   
0 1 1 0 0 1
Hs ( )dx = Hs ( )dx
A y x A 0 y 1 xy −1
Z  
0 1
=(y)|y|1−s Hs ( )dx,
A 1 x

where in the last equation we substituted x by xy −1 . This last expression only

depends on the divisor class of y. But it is also the difference of two functions, the
Fourier coefficient |d|−1/2 Es0 (y) and |d|−1/2 |y|s . Since  is ramified this can hold for
this integral only if it is zero. We conclude that Es0 (y) = |y|s . 

6. Poincaré Series
Lemma 6.1. For any Borel measure µ on Z(Fv )\G(Fv ) the following two proper-
ties are equivalent:
(i) µ is a left-invariant Haar-measure,
(ii) µ(Sk) = µ(S) for all k ∈ G(Ov ) and S ∈ B(Z(Fv )\G(Fv )), and µ(bS) = µ(S)
for all b ∈ Z(Fv )\B(Fv ) and S ∈ B(Z(Fv )\G(Fv )).
Proof. We will first prove that (i) implies (ii). It is clear that the second prop-
erty of (ii) holds. For every k ∈ G(Ov ) the measure µ(·k) is also a left-invariant
Haar measure, therefore there is a positive real number m(k) such that µ(Sk) =
m(k)µ(S) for all S ∈ B(Z(Fv )\G(Fv )). Since the function m is a continuous char-
acter on the compact group G(Ov ), it must be constant, hence the first property
in (ii) also holds.
In order to prove that (ii) implies (i) we first remark that the restriction of any
µ satisfying (ii) to B(Z(Fv )\G(Ov )) is a left-invariant Haar-measure of the group
Z(Ov )\G(Ov ), because it is right-invariant and the group is compact. This fact and
the Iwasawa decomposition implies that µ(gS) = µ(S) for all g ∈ Z(Fv )\G(Fv ) and
for all S which is a translate of a compact open group. But the latter generate the
σ-algebra of Borel sets, hence this property holds for each Borel set as well. 
Definition 6.2. Let dk = ⊗dkv be the Haar measure on K0 (1) with volume 1
on each component, let dx = ⊗dxv be the Haar measure on A such that Ov has
volume one, and let d∗ x = ⊗d∗ xv be the Haar measure such that Ov∗ has volume
one. Now define the measure dg on Z(A)\G(A) by the formula
Z Z Z Z  
y yx
f (g)dg = f( k)dkdxd∗ y.
Z(A)\G(A) A∗ A K0 (1) 0 1

Lemma 6.3. The measure dg is a left-invariant Haar measure on Z(A)\G(A).

Proof. The measure dg is a product of local measures which all satisfy (ii) of the
lemma above, since dg satisfies it. Hence they are left-invariant, and so does dg. 
14 AMBRUS PÁL

Definition 6.4. Because dg is a left-invariant Haar measure, it induces a measure

on Z(A)G(F )\G(A), which will be denoted by the same letter by abuse of notation.
Let ω be a grössencharacter of F . For any pair of automorphic forms φ, ψ with
central character ω, the integral
Z
φ(g)ψ(g)dg,
Z(A)G(F )\G(A)

if it is absolutely convergent, will be denoted by hφ, ψi, and will be called the
Peterson product of φ and ψ.
Definition 6.5. Let n, m be two effective divisors on X. Let τ : F \A → C be a
nontrivial additive character. Let H n,m denote the unique complex-valued function
on G(A) which satisfies the following condition: for all g ∈ G(A) which is of the
form  
a b
g= k,
0 c

with  
a b
∈ B(A), k ∈ K0 (n),
0 c

then
H n,m (g) = τ (b/c),

if (a/c) = md, and H n,m (g) = 0, otherwise.

Proposition 6.6. (i) The series
X
P n,m (g) = H n,m (γg),
γ∈Z(F )U (F )\G(F )

is uniformly absolute convergent and defines an automorphic form with trivial cen-
tral character and level n which is compactly supported as a function on the topo-
logical space Z(A)G(F )\G(A).
(ii) Let f be a cuspidal automorphic form of level n and of trivial central char-
acter. Then we have:

|m|−1 |d|−3/2 f ∗ (m)

hf, P n,m i = ,
[K0 (1) : K0 (n)]

where [K0 (1) : K0 (n)] is the index of K0 (n) in K0 (1).

Proof. By Proposition 4.6 it is sufficient to prove that H n,m is compactly supported
as a function on Z(A)U (F )\G(A) in order to prove that (i) holds, since P n,m is
K0 (n)-invariant on the right by definition. The support of this function is the image
of the product U (A)S(m)K0 (n), where the set S(m) is defined as
  
a 0
S(m) = ∈ B(A)|(a) = m .
0 1
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 15

The image of all these sets are compact modulo Z(A)U (F ), since S(m) and K0 (n)
are themselves compact, and because the image of U (A) factors through the com-
pact set U (F )\U (A). In order to show (ii), we first note that the function f ·H n,m :
Z(A)U (F )\G(A) → C has compact support, so the series
X
|f (g)H m,n (γg)|
γ∈Z(F )U (F )\G(F )

is uniformly convergent, has compact support, hence defines a function which is

Lebesgue-integrable by Proposition 4.6. Therefore by the definition of P n,m and
Proposition 4.7 we have that
Z
hf, P n,m i = f (g)P m,n (g)dg
Z(A)G(F )\G(A)
Z
= f (g)H m,n (g)dg
Z(A)U (F )\G(A)
Z Z Z  
y yx
= (f H m,n )( k)dkdxd∗ y
A ∗ A/F K0 (1) 0 1
Z Z Z  
y yx
= f( )τ (−yx)dkdxd∗ y
(y)=md yF \A K0 (n) 0 1
|md|−1
Z  
md x
= f( )τ (−x)dkdxd∗ y
[K0 (1) : K0 (n)] F \A 0 1
|m|−1 |d|−3/2 f ∗ (m)
= ,
[K0 (1) : K0 (n)]

where m also denotes an idéle whose divisor is m by abuse of notation. 

7. Estimates of the Fourier coefficients of Poincaré Series

Lemma 7.1. There is a constant C depending only on X such that for any line
bundle L ∈ Pic(X) we have:
deg(L)
r(L) = |{s ∈ H 0 (L, X)|the divisor (s) has only even coefficients}| ≤ Cq 2 .

Proof. A section s ∈ H 0 (L, X) is as above if and only if there is a unique effective

divisor D ∈ Div(X) such that 2D = (s). This implies that

r(L) ≤ q|{D ∈ Div(X)|D is effective, 2[D] = L}|.

But we can estimate as follows:

X 0
|{D ∈ Div(X)|D is effective, 2[D] = L}| ≤ q dim(H (M,X))

M∈Pic(X)
2M=C
deg(L)
≤|Pic0 (X)|q 1+ 2 .
16 AMBRUS PÁL

Lemma 7.2. Let L = Fqn ((t)) be a field of Laurent series, let D = Fqn [[t]] be
P∞ ring and let χ : L → C be the additive character defined by
its discrete valuation
the formula χ( i=−N ai ti ) = χ0 (a−1 ), where χ0 : Fqn → C is a nontrivial additive
character. Then for any a, b, and c ∈ D − {0}, for the sum

X ax−1 + bx
S(a, b, c) = χ( )
c
x∈(D/cD)∗

we have the estimate

n
|S(a.b, c)| ≤ 2q 2 v(c)+nmin(v(a),v(b),v(c)) ,

where v : L∗ → Z is the valuation.

Proof. We can write c as a product c = tm u, where u ∈ D∗ . As we have that
S(a, b, c) = S(u−1 a, u−1 b, tn ), we can assume that c = tm . Similarly, if a = a0 tr ,
b = b0 tr and c = c0 tr for some a0 , b0 , c0 ∈ D and positive r, then we have S(a, b, c) =
q nr S(a0 , b0 , c0 ), so we can assume that one of a and b, for example a is in D∗ . Then
there are three cases:
(i) c = t2m for some positive m. Choose a representative system S in the set
(D/t2m D)∗ for the factor set (D/tm D)∗ . As every element in (D/t2m D)∗ can be
written uniquely as a product y(1 + xtm ) for some y ∈ S and x ∈ D/tm D, we can
compute as follows:

X ax−1 + bx X X ay −1 (1 − xtm ) + by(1 + xtm )

χ( 2m
)= χ( )
t t2m
x∈(D/t2m D)∗ y∈S x∈D/tm D
X ay −1 + by X (by − ay −1 )x
= χ( ) χ( ),
t2m tm
y∈S x∈D/tm D

where we also used that the inverse of 1 + xtm is 1 − xtm . The sum

X (by − ay −1 )x
χ( )
tm
x∈D/tm D

can be nonzero only if by ≡ ay −1 mod tm D, in which case it is equal to q nm . Since

a is in D∗ there at most two y ∈ S which satisfies this congruence, hence the
estimate.
(ii) c = t2m+1 for some positive m. Choose a representative system S in the set
(D/t2m+1 D)∗ for the factor set (D/tm+1 D)∗ . Using the same argument as above
we have:
X ax−1 + bx X ay −1 + by
χ( ) = q nm χ( ).
t2m+1 t2m+1
x∈(D/t2m+1 D)∗ y∈S
by−ay −1 ∈tm D

Since our choice of S was arbitrary, we can assume that set of elements y ∈ S
which satisfy the congruence by ≡ ay −1 mod tm D, if it is not empty, is of the form
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 17

ei (1 + xtm ), where e1 , e2 are two fixed solutions of this congruence and x ∈ Fqn
arbitrary. As for such elements (1 + xtm )−1 = 1 − xtm + x2 t2m , we get

X ay −1 + by X X ae−1 m 2 2m
i (1 − xt + x t ) + bei (1 + xtm )
χ( ) = χ( )
y∈S
t2m+1 i=1,2
t2m+1
n x∈Fq
by−ay −1 ∈tm D

X ae−1
i + bei
X ae−1
i x
2
= χ( 2m+1
) χ( ),
i=1,2
t t
n x∈Fq

where terms involving xtm can be eliminated because of the congruence satisfied by
n
ei . The two sums we got are well known to have absolute value q 2 , which implies
our claim.
(iii) c = t. In this case the sum is identical to the one considered in [11] and our
estimate follows from the main result of that paper. 
Corollary 7.3. For any v ∈ |X|, a, b ∈ Ov and c ∈ Ov − {0} we have the estimate:

ax−1 + bx
Z
| τv ( )dx| ≤ 2max(|dv |v , |dv |−1 1/2
v )|c|v min(|a|v , |b|v , |c|v )
−1
.
Ov∗ c

Proof. Define the character τbv : Fv → C by the formula τbv (x) = τv (dv x). Then
we can write
−1
τb ( ax dv+bx
( R
Z
ax−1 + bx Ov∗ v c )dx, if |dv | ≤ 1,
τv ( )dx = R ad−1 −1
x +bd−1
x
Ov∗ c τb ( v v
)dx, if |dv | ≥ 1,
O∗ v c v

Since for any α, β ∈ Ov , γ ∈ Ov − {0} we have

αx−1 + βx αx−1 + βx
Z X
τbv ( )dx = |γ|v τbv ( )
Ov∗ γ γ
x∈(Ov /γOv )∗

and the sum on the right is of the type considered in Lemma 7.2, we have the claim
above as a corollary. 
Proposition 7.4. For any y ∈ A∗ and positive  there is a constant C(y, ) de-
pending only on y and  such that for any effective divisor n and m we have the
estimates:
|(P n,m )∗ (y)| ≤ C(y, )|m|−3/4− .

Proof. We will denote by n, m an idéle as well, whose divisor is n, m, respectively,

by abuse of notation. We will really estimate (P n,m )∗ (yd). We can assume that
(yd) is effective. The Fourier coefficient is given by the formula
Z  
n,m ∗ n,m y x
(P ) (yd) = P ( )τ (−x)dx,
F \A 0 1

where the measure dx on A is the usual one.

18 AMBRUS PÁL

As G(F ) has the pair-wise disjoint decomposition

    
[ u 0 [ [ 0 u 1 w
G(F ) = Z(F )U (F ) Z(F )U (F ) ,
∗
0 1 ∗
1 0 0 1
u∈F u∈F ,w∈F

we have
X Z  
uy ux
(P n,m )∗ (yd) = H n,m ( )τ (−x)dx
0 1
u∈F ∗ F \A
X Y Z  
0 uv
+ Hvn,m ( )τv (−x)dx.
∗ Fv yv x
u∈F v∈|X|

For the first term we have:

X Z   Z Z
uy ux X
| H n,m ( )τ (−x)dx| = | τ ((u − 1)x)dx| ≤ q dx.
∗ F \A 0 1 F \A F \A
u∈F (uy)=(md)

To estimate the second term we first prove a lemma.

Lemma 7.5. For any u, y ∈ Fv∗ we have
 R
Z   x∈yOv
τv (x)dx,
0 u
R
Hvn,m ( )τv (−x)dx = uy
|x|2v =| md |v
τv ( ux − x)dx,
Fv y x 
0,

if v(u) = v(y) + v(md) and v(n) = 0, v(y) + v(md) − v(u) is even and v(u) ≤
v(y) + v(md) − max(1, v(n)), and otherwise, respectively.
Proof. Write the integral as a sum of
Z  
0 u
Iv (u, y, m, n) = Hvn,m ( )τv (−x)dx
xy −1 ∈Ov y x

and Z  
0 u
Jv (u, y, m, n) = Hvn,m ( )τv (−x)dx.
−1
yx ∈πv Ov y x
For any u, x and y as above we have the product decompositions:
    
0 u u 0 0 1
=
y x 0 y 1 xy −1

and    uy  
0 u x u −1 0
= .
y x 0 x yx−1 1
Since in the first product the first matrix is in B(Fv ) and the second is in G(Ov )
when xy −1 ∈ Ov , we can write the first integral above as follows:
Z   
n,m u 0 0 1
Iv (u, y, m, n) = Hv ( )τv (−x)dx
xy −1 ∈Ov 0 y 1 xy −1
Z
= τv (x)dx,
x∈yOv
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 19

if v(u) = v(y) + v(md), and v(n) = 0 and Iv (u, y, m, n) = 0 otherwise. The second
condition for Iv (u, y, m, n) to be non-zero is necessary to have the second matrix in
the decomposition above to be in K0 (n)v .
Similarly in the second product the first matrix is in B(Fv ) and the second is in
G(Ov ) when yx−1 ∈ πv Ov , so we have:
Z  uy  
n,m x u −1 0
Jv (u, y, m, n) = Hv ( )τv (−x)dx
yx−1 ∈πv Ov 0 x yx−1 1
Z
u
= τv ( − x)dx,
2 uy
|x|v =| md |v x

if v(md) − v(u) − v(y) is even and max(1, v(n)) ≤ 1/2(v(md) + v(y) − v(u)), and
Jv (u, y, m, n) = 0 otherwise.
The conditions above make it impossible that Iv (u, y, m, n) and Jv (u, y, m, n) are
both non-zero, hence the lemma follows. 
Now we will estimate the infinite sum:
X Y Z  
0 uv
Hvn,m ( )τv (−x)dx.
∗ Fv y v x
u∈F v∈|X|

By Lemma 7.5 the term for u can only be non-zero, if the divisor ( ymd u ) is effective
and has only even coefficients. By Lemma 7.1 this implies that the number of
those u for which the term is non-zero is at most C|ymd|−1/2 for some constant C
depending only on X.
Now we will estimate the individual terms. Fix u and let π1 , π2 ∈ Fv such that
uy 1/2 ymd 1/2
|π1 |v = | md |v and |π2 |v = min(|d|v , |d|−1v )| u |v . It is clear that |π2 |v ≤ 1. Then
the integral

uπ2 π1−1 x−1 − π1 π2 x

Z Z
u uy 1/2
τv ( − x)dx = | |v τv ( )dx,
uy
|x|2v =| md |v x md Ov∗ π2

can be estimated using Corollary 7.3, since

umd 1/2 ymd 1/2
|uπ2 π1−1 |v = | | | | min(|d|v , |d|−1
v ) ≤ |m|v ≤ 1
y v u v
and
uy 1/2 ymd 1/2
|π1 π2 |v = | | | | min(|d|v , |d|−1
v ) ≤ |yd|v ≤ 1,
md v u v
as follows:
uy 3 1/4
Z
u
| τv ( − x)dx| ≤ 2| | |yd|v .
uy
|x|2v =| md |v x md v
Using this estimate for the factor of the term for those v ∈ |X| for which we have
−1/4
| ymd
u |v< 1 and the trivial estimate by |y|v | ymd
u |v otherwise, we get the estimate
Y Z  
0 uv ymd ymd −1/4
| Hvn,m ( )τv (−x)dx| ≤C 0 (y)2t(( u )) | |
Fv yv x u
v∈|X|
ymd
≤ C(y)2t(( u )) |m|−1/4 ,
20 AMBRUS PÁL

where we used that |u| = 1, moreover C(y) and C 0 (y) are constants depending
on y, and for any divisor g the number t(g) is defined as the number of prime
divisors of g. Since for every positive  there is a constant C() such that for any
effective divisor g we have the estimate 2t(g) ≤ C()|g|− , using our estimates on the
individual terms and the number of non-zero terms we get the inequality claimed
in Proposition 7.4. 
Proposition 7.6. For any y ∈ A∗ , closed point ∞ ∈ |X|, effective divisor m0
having disjoint support from ∞, and positive  there is a constant C(y, ∞, m0 , )
depending only on y, ∞, m0 and  such that for any effective divisor n and natural
number n we have the estimate:
|(P n,m )0 (y)| ≤ C(y, ∞, m0 , )|m|−1/2− ,
where m = m0 ∞n .
Proof. For the Fourier coefficient (P n,m )0 (y) we have
X Z  
uy ux
(P n,m )0 (y) = H n,m ( )dx
0 1
u∈F ∗ F \A
X Y Z  
0 uv
+ Hvn,m ( )dx.
∗ Fv yv x
u∈F v∈|X|

The first term is zero. To estimate the second term we need the following lemma
whose proof is same as the proof of Lemma 7.5.
Lemma 7.7. For any u, y ∈ Fv∗ we have

Z   |y|v ,
0 u
R
n,m u
Hv ( )dx = |x| 2 =| uy | τv ( x )dx,
Fv y x v md v
0,


if v(u) = v(y) + v(md) and v(n) = 0, v(y) + v(md) − v(u) is even and v(u) ≤
v(y) + v(md) − max(1, v(n)), and otherwise, respectively. 
Now we will estimate for all u ∈ F ∗ the infinite product:
Y Z  
n,m 0 uv
Hv ( )dx.
Fv yv x
v∈|X|

The product can be non-zero only if the divisor ( ymd u ) is effective and has only even
coefficients, which we will assume about u. For every v ∈ |X| choose a πv ∈ Fv
uy 1/2
such that |πv |v = | md |v . Then we have the estimate
Z Z
u uy 1/2 uy 1/2 yd 1/2
| τv ( )dx| = | |v | τv (uπv−1 x)dx| ≤ | |v · min(| |v , 1),
2 uy
|x|v =| md |v x md ∗
Ov md um
where we used the explicit expression for the second integral in the estimate. Using
this estimate for those v ∈ |X| for which we have | ymd
u |v < 1 and the trivial estimate
ym 1/2
by |y|v | ud |v otherwise, we get the estimate
Y Z  
0 uv Y yd 1/2
| n,m
Hv ( )dx| ≤ C(y)|m|−1/2 min(| |v , 1),
Fv y v x um
v∈|X| v∈|X|
 THETA SERIES, EISENSTEIN SERIES AND POINCARÉ SERIES 21

where C(y) is a constant depending on y. But

Y yd 1/2 Y yd 1/2 Y yd 1/2
min(| |v , 1) ≤ min(| | , 1) ≤ | | = C(y, m0 )|u|1/2
∞ ,
um um0 v um0 v
v∈|X| v∈|X|−{∞} v∈|X|−{∞}

where C(y, m0 ) is a constant depending on y and m0 . By Lemma 7.1 for all natural
1/2
number l the number of such u which satisfies the assumptions above and |u|∞ =
q − deg(∞)l = q −ml , is at most C 0 (y, m0 )q ml , where C 0 (y, m0 ) is a constant depending
on y and m0 . Also this set is empty if 2l ≥ n + C 00 (y, m0 ) or 2l ≤ −n − C 00 (y, m0 ),
where C 00 (y, m0 ) is a constant depending on y and m0 , so we get

|(P n,m )0 (y)| ≤ |m|−1/2 C(y)C(y, m0 )C 0 (y, m0 )(n + C 00 (y, m0 )). 

References
1. W. Casselman, On Some Results of Atkin and Lehmer, Math. Ann. 201 (1973), 301–314.
2. C. W. Curtis and I. Reiner, Methods of Representation Theory, vol. I, John Wiley&Sons, Inc.,
New York, 1981.
3. P. Deligne, Les constantes des equations fonctionnelles des fonctions L, Lecture Notes in
Math., vol. 349, Springer, Berlin-Heidelberg-New York, 1973, pp. 501–597.
4. E.-U. Gekeler and M. Reversat, Jacobians of Drinfeld Modular Curves, J. reine angew. Math.
476 (1996), 27–93.
5. S. S. Gelbart, Automorphic Forms on Adele Groups, Princeton University Press, Princeton,
1975.
6. R. Godement and H. Jacquet, Zeta Functions of Simple Algebras, Springer, Berlin-Heidelberg-
New York, 1972.
7. B. R. Gross, Heights and the Special Values of L-series, Number Theory (Montreal, Que.,
1985), Amer. Math. Soc., Providence, 1987, pp. 115-187.
8. B. H. Gross and D. Zagier, Heegner Points and Derivatives of L-series, Invent. Math. 84
(1986), 225–320.
9. G. Harder, Chevalley Groups over Function Fields and Automorphic Forms, Ann. Math. 100
(1974), 249–306.
10. G. Harder, W. C. W. Li and j. R. Weisinger, Dimensions of Spaces of Cusp Forms over
Function Fields, J. Reine Angew. Math. 319 (1980), 73–103.
11. H. Jacquet and R.P. Laglands, Automorphic Forms on GL(2), Springer, Berlin-Heidelberg-
New York, 1970.
12. H. Jacquet, I. I. Piatevski-Shapiro, and J. Shalika, Conducteur des représentations du group
linéare, Math. Ann. 256 (1981), 199–214.
13. S. Lang, Algebraic Number Theory, Addison-Wesley Publishing, Inc., Reading-London, 1970.
14. W. C. W. Li, Eisenstein Series and Decomposition Theory over Function Fields, Math. Ann.
240 (1979), 115–139.
15. H.-G. Rück, Theta Series of Imaginary Quadratic Function Fields, Manuscripta Math. 88
(1995), 385–407.
16. H.-G. Rück, Poincaré Series of Drinfeld Type, Arch. Math. 68 (1997), 190–201.
17. H.-G. Rück, L-series of Automorphic Cusp Forms of Drinfeld Type, Drinfeld Modules, Mod-
ular Schemes and Applications (Alden-Biesen, 1996), World Sci. Publishing, River Edge, NJ,
1997, pp. 311–329.
18. H.-G. Rück and U. Tipp, Heegner Points and L-series of Automorphic Cusp Forms of Drinfeld
Type, Doc. Math. 5 (2000), 365–444.
19. J.-P. Serre, Corps Locaux, Hermann, Paris, 1971.
20. J.-P. Serre, Arbres, Amalgames, SL2 , Société Mathématique de France, Paris, 1977.
21. A. Weil, On Some Exponential Sums, Collected Papers, vol. 1, 1979, pp. 386–389.
22. A. Weil, Dirichlet Series and Automorphic Forms, Springer, Berlin-Heidelberg-New York,
1971.
23. A. Weil, Basic Number Theory, Springer, Berlin-Heidelberg-New York, 1973.
24. K. S. Williams, The Kloosterman Sum Revisited, Canad. Math. Bull. 16 (1973), 363–365.
25. S.-W. Zhang, Heights of Heegner Points on Shimura Curves, Ann. Math. 153 (2001), 27–147.