On analytic properties of L -functions attached to cusp forms on the unitary group of degree two

On analytic properties of L -functions attached to cusp forms on the unitary group of degree two

Tomonori KADOTA (Last Updated: 12 Dec, 2005)

Abstract

Let f be a holomorphic cusp form on U(1,1). In this paper, we study the standardL-function offand show its functional equation under certain conditions.

2000 Mathematics Subject Classification: Primary 11F11; Secondary 22E35.

Key words: Automorphic form, L-function, Rankin-Selberg convolution, Whittaker function.

1 Introduction

Let f be a holomorphic cusp form on U(1,1). In this paper, we study the standard L-function off and show its functional equation under certain conditions. The proof is based on the method of Shimura ([Sh2]).

We explain our results in a classical formulation in the simplest case. Let K be an imaginary quadratic field of discriminantDwith|D|>4. For simplicity, we assume that Dis odd and the class number ofK is 1. Let be a positive integer divisible bywK, the number of roots of unity inK. Let

f(z) = ∞ m=1

c(m)e^2πimz

be a holomorphic cusp form of weight −1 and character D

∗

on Γ0(|D|). The L- function off we are concerned with is given by

L(f;s) = ζ(2s)

a

c(Na)αNa^−(s+−1)

×

p|D

(1−p^−2s)

1−c(p)Π_pp^−s−+1₋₁ ,

wherea= (α) runs over the nonzero integral ideals ofK and, forp|D, Πpis a generator of the prime ideal of K dividingp.

Theorem Suppose thatf is a Hecke eigenform withc(1)= 0and an eigenform of Atkin- Lehner operator at each p|D. Set

L^∗(f;s) = (2π)^−2s|D|^sΓ(s+ 1)Γ(s+−1)L(f;s),

where Γ(s)is the gamma function. Then L^∗(f;s) is continued to an entire function of s onC and satisfies

L^∗(f;s) =L^∗(f; 1−s).

The result in a general case is stated in Section 4 in an adelic formulation. The main ingre- dients of the proofs are the Rankin-Selberg convolution and the local theory of Whittaker functions.

The paper is organized as follows. In Section 2, we prepare some notations used throughout this paper. In Section 3, we recall the definitions of cusp forms, Hecke opera- tors, Atkin-Lehner operators, automorphicL-functions, theta series and Eisenstein series on U(1,1). We also calculate the Fourier coefficients of the Eisenstein series. It is to be noted that a similar calculation was made by Shimura ([Sh2]) in a much more general situation. The main results of this paper are given in Section 4. In Section 5, we study the local and global Whittaker functions. Using the Rankin-Selberg convolution together with the results in Section 5, we give proofs of our results in Section 6. We state the classical interpretations of cusp forms in Section 7. Finally, in Section 8, we present an example.

2 Preliminaries

2.1 Notations

As usual,Z, Q, R andCdenote the ring of rational integers, the rational number field, the real number field and the complex number field respectively. We writeifor√

−1. For a ring R, R^× denotes the group of all invertible elements ofR. Let Z+ and R+ denote the set of positive rational integers and that of positive real numbers respectively. For a set S, charS stands for the characteristic function of S. For a primev of Q, Qv denotes the completion ofQatv. For a finite primep,Zpdenotes thep-adic integer ring. We put Zf =

p<∞

Zp. LetQ_Adenote the adele ring ofQ. Forx∈Q_A, letxv be thev-component ofxfor each primev. For a primev,| · |_v stands for the absolute value ofQ^×_v. Forv=∞,

we often write | · |for | · |_∞. Forx= (xv)v ∈Q^×_A, let|x|_A=

v≤∞

|xv|_v be the idele norm of x. For a finite prime p, we normalize the additive valuation ordp : Q^×_p → Z so that ordpp= 1. In this paper, we fix an imaginary quadratic field K of discriminantD with integer ring OK. We only consider the case of |D| > 4. Then we have wK = 2, where wK is the number of roots of unity in K. Denote byσ the nontrivial automorphism of K/Q. Forz ∈K, let Tr(z) =z+z^σ and N(z) =zz^σ. The complex conjugate of z∈C is denoted by z. For a prime v of Q, letKv = K⊗QQv. For a finite prime p, we put OK,p =OK ⊗ZZp. We set OK,f =

p<∞

OK,p. We denote by KA andKA,f the adele ring of K and its finite part respectively. Fora= (av)v ∈K_A^×, put a _A=

v≤∞

av v, where av v=|N(av)|_v. We putK¹={t∈K^×; N(t) = 1}andO_K,f¹ =K_A,f¹ ∩ O^×_K,f =

p<∞

O¹_K,p, whereO_K,p¹ =K_p¹∩O_K,p^× . Whenpramifies inK/Q(namelyp|D), we fix a prime element Πpof Kp. Whenpsplits inK/Q, we fix an identification betweenKpand Qp⊕Qp and put Πp,1= (p,1) and Πp,2= (1, p).

2.2 Characters

Let X be the set of Hecke charactersχ of K satisfying χ|_Q^×_A =ω, where ω denotes the quadratic Hecke character of Q corresponding toK/Q by class field theory. Forχ∈ X, let w∞(χ) be the integer such that χ(z∞) = (z∞/|z∞|)^w∞(χ) for z∞ ∈ C^×. We fix an elementχ0∈ X such thatw∞(χ0) =−1. Letψv be the additive character ofQv given by

ψv(x) =

e^2πix (v =∞), e^−2πi{x}p (v =p),

where {x}_p denotes the fractional part of x ∈ Qp. Then ψ =

v≤∞

ψv is a nontrivial character ofQ\QA. We putψKv =ψv◦Tr.

2.3 Unitary group

LetH=U(T) be the unitary group ofT =

1

−1

. Namely HQ= h∈GL2(K); ^th^σT h=T

.

We put

n(x) = 1 x

1

, d(y) = y^σ

y⁻¹

, n(z) = 1

z 1 and S =

1

−1

(x∈Q, y∈K^×, z∈Q).

Define subgroups of H by

N ={n(x);x∈Q}, Z ={a·I;a∈K¹}, P ={n(x)d(y);x∈Q, y∈K^×},

where I is the identity matrix of degree 2. We denote byHQ(resp. Hv) the group of the Q-rational (resp. Qv-rational) points ofH. LetHA be the adele group of H and HA,f

the finite part of HA. For a finite primepof Q, letUp=Hp∩GL2(OK,p) and U0(D)p= a b

c d

∈ Up; c∈DOK,p

. Note thatU0(D)p=Up unlessp|D. We putUf =

p<∞

Up

and U0(D)f =

p<∞

U0(D)p. Let U∞ = {h∈H∞;hi=i}, where hz = az+b cz+d for h=

a b c d

∈H∞ andz ∈H={z∈C; Im(z)>0}. Forh∈H∞ andz ∈H, we define the automorphic factors by j(h, z) = cz+d and J(h, z) = (deth)⁻¹j(h, z). Let χ0,p be the character of U0(D)p given by

χ0,p

a b c d

=

χ0,p(a) (c∈pOK,p),

χ0,p(c) (c∈ OK,p−pOK,p).

Then χ0=

p<∞

χ0,p defines a character ofU0(D)f.

2.4 Measures

Let dxp be the Haar measure on Q_p normalized by

Zp

dxp = 1. Let dx∞ be the usual Lebesgue measure onR. Thendx=

v≤∞

dxvis the Haar measure onQ_Awith

Q\QA

dx= 1. Let d^×xp be the Haar measure on Q^×_p normalized by

Z^×p

d^×xp = 1. Let d^×x∞ =

|x∞|⁻¹dx∞. Thend^×x=

v≤∞

d^×xv is the Haar measure onQ^×_A. For each prime v of Q, letdyvbe the Haar measure onKvself-dual with respect to the pairing (x, y)→ψKv(xy^σ).

Note that

OK,p

dyp = |D|^1/2_p if v = p < ∞, and that dy∞ is twice the usual Lebesgue

measure on C. Thendy =

v≤∞

dyv is the Haar measure onKA with

K\KA

dy = 1. Let d^×y =

v≤∞

d^×yv be the Haar measure on K_A^×, where d^×yv is the Haar measure on K_v^× normalized by

O_K,p^× d^×yp= 1 if v=p <∞, andd^×y∞= 2⁻¹N(y∞)⁻¹dy∞. For a prime v of Q, we normalize the Haar measure dhv on Hv by

Hv

f(hv)dhv=

Qv

dxv

Kv^×

d^×yv

Uv

duv yv −1

v f(n(xv)d(yv)uv), where dup and du∞ are the Haar measures on Up and U∞ normalized by

U0(D)p

dup =

U∞

du∞= 1, respectively. Thendh=

v≤∞

dhv is the Haar measure onHA.

3 Definitions

3.1 Cusp forms

Let∈Z+withwK |. A smooth functionfonHQ\HAis called acusp form onU0(D)f

of weight −1with characterχ0if the following conditions (1)–(4) are satisfied.

(1) f(hufu∞) =χ0(uf)J(u∞, i)¹⁻f(h) (h∈HA, uf ∈ U0(D)f, u∞∈ U∞).

(2) For every hf ∈ HA,f, the function fdm,hf : H h∞i → J(h∞, i)⁻¹f(h∞hf) (h∞∈H∞) is holomorphic.

(3) For everyhf ∈HA,f,fdm,hf is holomorphic ati∞. (4)

Q\QA

f(n(x)h)dx= 0 (h∈HA).

We denote by S−1(D, χ0) the space of such functions. LetYbe the set of characters Ω of K_A¹/K¹ satisfying Ω|O¹_K,f =1and Ω(z∞) =z_∞ for z∞∈C¹. For Ω∈ Y, we put

S−1(D, χ0;χ0Ω) = f∈S−1(D, χ0);f(th) = (χ0Ω)(t)f(h) (t∈K_A¹, h∈HA) . Then S−1(D, χ0) =

Ω∈Y

S−1(D, χ0;χ0Ω).

For f ∈S−1(D, χ0;χ0Ω), we define the global Whittaker functionWf attached to f by

Wf(h) =

Q\QA

ψ(−x)f(n(x)h)dx (h∈HA).

It is easily seen that

Wf(tn(x)hufu∞) = (χ0Ω)(t)ψ(x)χ0(uf)J(u∞, i)¹⁻Wf(h) for t∈K_A¹, x∈Q_A, uf ∈ U0(D)f andu∞∈ U∞.

3.2 Hecke operators and Atkin-Lehner operators

Letf ∈S−1(D, χ0;χ0Ω). For each finite primep, we defineHecke operatorsas follows.

(i) Suppose thatpis inert inK/Q. Then we put Tpf(h) =−f(hd(p⁻¹))−

x∈Z^×p/pZp

f(hn(p⁻¹x))−

y∈Zp/p²Zp

f(hn(y)d(p)).

(ii) Suppose that pramifies in K/Q. Then we put Tpf(h) =χ0,p(Πp)⁻¹

x∈Zp/pZp

f(hn(Dx)d(Π⁻¹_p )) +χ0,p(Πp)

y∈Zp/pZp

f(hn(y)d(Πp)).

(iii) Suppose thatpsplits in K/Q. Then we put Tp,1f(h) = χ0,p(Πp,1)⁻¹

⎧⎨

⎩f(hd(Π⁻¹p,1)) +

x∈Zp/pZp

f(hn(x)d(Πp,2))

⎫⎬

⎭,

Tp,2f(h) = χ0,p(Πp,2)⁻¹

⎧⎨

⎩f(hd(Π⁻¹p,2)) +

x∈Zp/pZp

f(hn(x)d(Πp,1))

⎫⎬

⎭.

Remark 3.1 Note thatS−1(D, χ0;χ0Ω) is invariant under Hecke operators, andTp,2f = Ω(Πp,2/Πp,1)Tp,1f for psplit inK/Q.

We say that f ∈S−1(D, χ0;χ0Ω) is a Hecke eigenform with eigenvalues {Λp} (Λp = (Λp,1,Λp,2)∈C² ifp splits inK/Q and Λp ∈C ifp does not split inK/Q) if, for every p < ∞, Tpf = Λpf in the non-split case and Tp,jf = Λp,jf (j = 1,2) in the split case.

Note that Λp,2= Ω(Πp,2/Πp,1)Λp,1in the split case.

For each prime factor p of D, we put wD,p =

√ D⁻¹

√D

∈ Hp. We define the Atkin-Lehner operator FD,p by

(FD,pf)(h) =f(hwD,p)

forf ∈S−1(D, χ0;χ0Ω) (cf. [A-L]). Thenf→FD,pfdefines an involution ofS−1(D, χ0;χ0Ω) commuting with Hecke operators and the eigenvalues ofFD,p are{±1}.

3.3 L -function

Letf ∈S−1(D, χ0;χ0Ω) be a Hecke eigenform with eigenvalues{Λp}. Letkbe a positive integer withwK |k andk≥, and let Ξ a Hecke character ofK satisfying

Ξ|_OK,f^× =1 (3.1)

and

Ξ(z∞) = (z∞/|z∞|)^k− (3.2) for z∞ ∈ C^×. Here 1 is the trivial character. We define the automorphic L-function L(f,Ξ;s) by

L(f,Ξ;s) =

p<∞

Lp(f,Ξp;s) withs∈C. Here the local factorLp(f,Ξ_p;s) is given as follows:

Lp(f,Ξp;s)

=

⎧⎪

⎪⎪

⎨

⎪⎪

⎪⎩

(1 + (1−p−Λp)Ξp(p)p^−2s−1+ Ξp(p)²p^−4s)⁻¹ (p is inert inK/Q),

j=1,2

1−Λp,jΞp(Πp,j)p^−s−1/2+ Ω(Πp,j/Π^σ_p,j)Ξp(Πp,j)²p^−2s−1

(p splits inK/Q), 1−ΛpΞp(Πp)p^−s−1/2+ Ξp(Πp)²p^−2s−1

(p ramifies inK/Q).

3.4 Metaplectic representation and theta series

Letϕv be a Schwartz-Bruhat function onKv. Let

ϕv(z) =

Kv

ψKv(y_v^σz)ϕv(yv)dyv

be the Fourier transform of ϕv. We define the Weil constant λK,v(ψv) by

Kv

ϕv(zv)ψv(N(zv))dzv =λK,v(ψv)

Kv

ϕv(zv)ψv(−N(zv))dzv

(cf. [We2]). The following facts are well-known.

Lemma 3.2

(1) λK,v(ψv)²=ωv(−1)for every v.

(2) For a finite primep ofQ, we have

λK,p(ψp) =

⎧⎪

⎨

⎪⎩

1 (pD),

√p^ordpD

Z^×p

ωp(p^−ordpDt)ψp(p^−ordpDt)dt (p|D).

(3) λK,∞(ψ∞) =i.

(4)

v≤∞

λK,v(ψv) = 1.

LetS(KA) be the space of Schwartz-Bruhat functions onKA. Letχ1 be an element of X such that w∞(χ1) = 2k+ 1 and χ⁻¹0 χ1|_O^×_K,f = 1. It is known that there exists a smooth representation M^Tχ1 of HA onS(KA) determined by

M^Tχ1(d(a))ϕ(X) = χ1(a)⁻¹ a ^1/2_A ϕ(aX) (a∈K_A^×), M^Tχ1(n(b))ϕ(X) = ψ(bN(X))ϕ(X) (b∈Q_A),

M^Tχ1(Sv)ϕ(X) = λK,v(ψv)

Kv

ψKv(Y_v^σXv)ϕ(YvX^(v))dYv (v ≤ ∞).

Here Sv ∈Hv and X^(v) =

v=v

Xv. We call M^Tχ1 the metaplectic representation of HA. Letϕ0=

v≤∞ϕ0,v∈ S(KA), where

ϕ0,p(Xp) = char_OK,p(Xp), ϕ0,∞(X∞) = X_∞^k exp(−2π|X∞|²).

It is known that

M^Tχ1(uv)ϕ0,v=

χ0,p(up)ϕ0,p (v=p <∞, up∈ U0(D)p), J(u∞, i)^−k−1ϕ0,∞ (v=∞, u∞∈ U∞).

We define atheta seriesby

θχ1(h) =

X∈K

M^Tχ1(h)ϕ0(X).

Then θχ1∈Sk+1(D, χ0).

3.5 Eisenstein series

In what follows, we fix an Ω∈ Y. Letξ be the Hecke character ofK given by

ξ(z) = (χ0χ⁻¹1 Ξ)(z)Ω(z/z^σ) (z∈K_A^×). (3.3) Note that ξ|_O^×_K,f =1 and ξ(z∞) = (z∞/|z∞|)^−k+−2 for z∞ ∈C^×. For s∈C, we define an Eisenstein seriesby

Ek−+2(h,Ξ;s) =

γ∈PQ\HQ

φk−+2(γh;s) (k≥), where

φk−+2(n(b)d(a)ufu∞;s) =ξ(a) a ^s_AJ(u∞, i)^−k+−2

for a ∈ K_A^×, b ∈ Q_A, uf ∈ Uf and u∞ ∈ U∞. The series Ek−+2(h,Ξ;s) converges absolutely and uniformly for (h, s)∈C×C, where C (resp. C) is any compact subset of HA(resp. of{s∈C; Re(s)>1}). Note that

Ek−+2(γthufu∞,Ξ;s) =ξ(t^σ)J(u∞, i)^−k+−2Ek−+2(h,Ξ;s) for γ∈HQ, t∈K_A¹,uf ∈ Uf andu∞∈ U∞.

Proposition 3.3 LetPr(s) = r j=0

(s+r−j). Put

E_k−+2^∗ (h,Ξ;s) =π^−sΓ(s)ζ(2s)P(k−)/2(s)Ek−+2(h,Ξ;s).

ThenE_k−+2^∗ (h,Ξ;s)is continued to an entire function ofsonC, and satisfies a functional equation

E_k−+2^∗ (h,Ξ;s) =E_k−+2^∗ (h,Ξ; 1−s).

Proof. We first prepare for the proof of this proposition. For simplicity, we writeκ=k−.

Put G(s) =π^−s/2Γ(s/2)ζ(s). The function G(s) is holomorphic on C except fors = 0 ands= 1, and has a functional equationG(s) =G(1−s).

By the definition of ξ, we have the following. Since Π²_p ∈ pO^×_K,p for p | D, we have ξp(Π_p)² = ξp(p). The equation 1 = ξ(p) = ξ∞(p)ξp(p) = ξp(p) implies that ξp(p) = 1 for all p. Moreover we have ξ(a^σ)⁻¹ = ξ(a) for a ∈ K_A^× since ξ(N(a)) = Ξ(N(a)) and N(a)∈Q^×_A=Q^×R₊

p<∞

Z^×_p.

We define the classical Whittaker functionWν,μ(z) by Wν,μ(z) = e^−z/2z^μ+1/2

Γ(μ−ν+ 1/2) _∞

0

e^−ztt^{μ−ν−1/2}(1 +t)^μ+ν−1/2dt

for ν, μ (Re(μ−ν+ 1/2)>0) and z ∈C(|arg(z)|< π). It is well-known thatWν,μ(z) is continued to an entire function of (ν, μ) on C², and satisfies a functional equation Wν,μ(z) =Wν,−μ(z).

We now calculate the Fourier coefficients ofEκ+2(h,Ξ;s):

Eκ+2(h,Ξ;s) =

m∈Q

E_κ+2^(m)(h,Ξ;s), where

Eκ+2^(m)(h,Ξ;s) =

Q\QA

ψ(−mx)Eκ+2(n(x)h,Ξ;s)dx.

From now on, we fix anh=n(b)d(a)ufu∞ ∈HA (b∈Q_A,a∈K_A^×, uf ∈ Uf, u∞∈ U∞).

By the Bruhat decompositionHQ=PQ∪PQSNQ, we have Eκ+2^(m)(h,Ξ;s) =

Q\QA

ψ(−mx)

φκ+2(n(x)h;s) +

r∈Q

φκ+2(Sn(x+r)h;s)

dx

= φκ+2(h;s)

Q\QA

ψ(−mx)dx +

Q\QA

r∈Q

ψ(mr)ψ(−m(x+r))φκ+2(Sn(x+r)h;s)dx

= δm,0φκ+2(h;s) +I^(m)(h;s), where δa,bis the Kronecker’s delta and

I^(m)(h;s) =

QA

ψ(−mx)φκ+2(Sn(x)h;s)dx.

Since

I^(m)(n(b)d(a)ufu∞;s) =ψ(mb)J(u∞, i)^−κ−2I^(m)(d(a);s)

for b ∈ Q_A, a ∈ K_A^×, uf ∈ Uf and u∞ ∈ U∞, we only have to consider I^(m)(d(a);s) (a∈K_A^×). DecomposeI^(m)(d(a);s) asI^(m)(d(a);s) =

v≤∞

Iv^(m)(d(av);s),

Iv^(m)(d(av);s) =

Qv

ψv(−mx)φκ+2(Svn(x)vd(av);s)dx.

Form∈QandM ∈Z, putμ= ordpmandRp(m, M;s) =

M+μ

n=0

p^(−2s+1)nat a finite prime p.

Suppose that p is inert. For ap ∈ Kp^×, put ordKap = Aif ap ∈ p^AO^×_K,p. It is easily seen that

I_p^(m)(d(ap);s) =

p^2AZp

ψp(−mx)ξp(p^−A)p^−As

pdx +

Qp−p^2AZp

ψp(−mx)ξp(p^A−ordpx)p^A−ordpxs

pdx

= ξp(p)^−Ap^2As

p^2AZp

ψp(−mx)dx +ξp(p)^Ap^−2As

2A−1

n=−∞

ξp(p)⁻ⁿp^(2s−1)n

Z^×p

ψp(−mpⁿx)dx.

If m= 0, we have I_p⁽⁰⁾(d(ap);s)

= ξp(p)^−Ap^2A(s−1)+ξp(p)^A(1−p⁻¹)p^−2As

2A−1

n=−∞

ξp(p)⁻ⁿp^(2s−1)n

= ξp(p)^−Ap^2A(s−1)

1 +ξp(p)^2A(1−p⁻¹)p^{−2A(2s−1)} ∞ n=−2A+1

ξp(p)ⁿp^(−2s+1)n

= ξp(p)^−Ap^2A(s−1)

1 +ξp(p)(1−p⁻¹)p^−2s+1 ∞ n=0

ξp(p)ⁿp^(−2s+1)n

= ξp(p)^−Ap^2A(s−1) 1 +ξp(p)(1−p⁻¹)p^−2s+1

1−ξp(p)p^−2s+1−1!

= ξp(p)^−Ap^2A(s−1)

1−ξp(p)p^−2s+1₋₁

1−ξp(p)p^−2s . If m∈p^−2AZ_p, we obtain

I_p^(m)(d(ap);s) = 0.

If m∈p^−2AZp, we have I_p^(m)(d(ap);s)

= ξp(p)^−Ap^2A(s−1)

−ξp(p)^A+μ+1p−2As+(−2s+1)(μ+1)

1−(1−p⁻¹)

2A+μ

n=0

ξp(p)⁻ⁿp^(2s−1)n

= ξp(p)^−Ap^2A(s−1)

"

1−ξp(p)^2A+μ+1p(−2s+1)(2A+μ+1)

×

1−(1−p⁻¹)1−ξp(p)^{−2A−μ−1}p(2s−1)(2A+μ+1)

1−ξp(p)⁻¹p^2s−1

#

= ξp(p)^−Ap^{2A(s−1)−1}

1−ξp(p)⁻¹p^2s−1₋₁

×

1−ξp(p)⁻¹p^2s 1−ξp(p)^2A+μ+1p(−2s+1)(2A+μ+1)

= ξp(p)^−Ap^2A(s−1)

1−ξp(p)p^−2s+1₋₁

×

1−ξp(p)p^−2s 1−ξp(p)^2A+μ+1p(−2s+1)(2A+μ+1)

= ξp(p)^−Ap^2A(s−1)

1−ξp(p)p^−2s

Rp(m,2A;s).

Hence we see that

I_p^(m)(d(ap);s) =

⎧⎪

⎨

⎪⎩

p^2A(s−1)(1−p^−2s+1)⁻¹(1−p^−2s) (m= 0), p^2A(s−1)(1−p^−2s)Rp(m,2A;s) (m∈p^−2AZp),

0 (m∈p^−2AZp),

where A= ord_Kap andμ= ord_pm.

Suppose that p ramifies. Forap ∈Kp^×, put A= ordKap if ap ∈Π^A_pO_K,p^× . It is easily seen that

I_p^(m)(d(ap);s) =

p^AZp

ψp(−mx)ξp(Π^−A_p )Π^−A_p ^s

pdx +

Qp−p^AZp

ψp(−mx)ξp(Π^{A−2 ord}_p ^px)Π^{A−2 ord}_p ^pxs

pdx

= ξp(Πp)^−Ap^As

p^AZp

ψp(−mx)dx +ξp(Πp)^Ap^−As

A−1

n=−∞

ξp(Πp)⁻²ⁿp^(2s−1)n

Z^×p

ψp(−mpⁿx)dx.

If m= 0, we have I_p⁽⁰⁾(d(ap);s)

= ξp(Πp)^−Ap^A(s−1)+ (1−p⁻¹)ξp(Πp)^Ap^−As ∞ n=−A+1

ξp(Πp)²ⁿp^(−2s+1)n

= ξp(Πp)^−Ap^A(s−1)

1 + (1−p⁻¹)ξp(Πp)²p^−2s+1 ∞ n=0

ξp(Πp)²ⁿp^(−2s+1)n

= ξp(Πp)^−Ap^A(s−1)

1−ξp(Πp)²p^−2s+1−1

1−ξp(Πp)²p^−2s . If m∈p^−AZp, we obtain

I_p^(m)(d(ap);s) = 0.

If m∈p^−AZ_p, we have Ip^(m)(d(ap);s)

= ξp(Πp)^−Ap^A(s−1)

−ξp(Πp)^A+2μ+2p−As+(−2s+1)(μ+1)

1−(1−p⁻¹)

A+μ

n=0

ξp(Πp)⁻²ⁿp^(2s−1)n

= ξp(Πp)^−Ap^A(s−1)

"

1−ξp(Πp)^2A+2μ+2p(−2s+1)(A+μ+1)

×

1−(1−p⁻¹)1−ξp(Πp)^−2(A+μ+1)p(2s−1)(A+μ+1)

1−ξp(Πp)⁻²p^2s−1

#

= ξp(Πp)^−Ap^A(s−1)

1−ξp(Πp)²p^−2s

×

1−ξp(Π_p)²p^−2s+1₋₁

1−ξp(Π_p)^2(A+μ+1)p(−2s+1)(A+μ+1)

= ξp(Πp)^−Ap^A(s−1)

1−ξp(Πp)²p^−2s

Rp(m, A;s).

Hence we see that I_p^(m)(d(ap);s) =

⎧⎪

⎨

⎪⎩

ξp(Π_p)^−Ap^A(s−1)(1−p^−2s+1)⁻¹(1−p^−2s) (m= 0), ξp(Πp)^−Ap^A(s−1)(1−p^−2s)Rp(m, A;s) (m∈p^−AZp),

0 (m∈p^−AZp),

where A= ord_Kap andμ= ord_pm.

Suppose that psplits. Forap∈K_p^×, put A= ord_Kap=a1+a2 ifap∈Π^a_p,1¹Π^a_p,2²O^×K,p. It is easily seen that

Ip^(m)(d(ap);s)

= ξp(Πp,1)^−a2ξp(Πp,2)^−a1p^As

p^AZp

ψp(−mx)dx +ξp(Πp,1)^a1ξp(Πp,2)^a2p^−As

A−1

n=−∞

ξp(p)⁻ⁿp^(2s−1)n

Z^×p

ψp(−mpⁿx)dx.

If m= 0, we have I_p⁽⁰⁾(d(ap);s)

= ξp(Πp,1)^−a2ξp(Πp,2)^−a1p^A(s−1)

×

1 + (1−p⁻¹)ξp(p)^Ap^−A(2s−1) ∞ n=−A+1

ξp(p)ⁿp^(−2s+1)n

= ξp(Πp,1)^−a2ξp(Πp,2)^−a1p^A(s−1)

1 + (1−p⁻¹)ξp(p)p^−2s+1 ∞ n=0

ξp(p)ⁿp^(−2s+1)n

= ξp(Πp,1)^−a2ξp(Πp,2)^−a1p^A(s−1)

1−ξp(p)p^−2s+1−1

1−ξp(p)p^−2s . If m∈p^−AZ_p, we obtain

I_p^(m)(d(ap);s) = 0.

If m∈p^−AZ_p, we have I_p^(m)(d(ap);s)

= ξp(Π_p,1)^−a2ξp(Π_p,2)^−a1p^A(s−1)

×

"

1−ξp(p)^A+μ+1p(−2s+1)(A+μ+1)

1−(1−p⁻¹)

A+μ

n=0

ξp(p)⁻ⁿp^(2s−1)n

#

= ξp(Πp,1)^−a2ξp(Πp,2)^−a1p^A(s−1)

1−ξp(p)p^−2s

×

1−ξp(p)p^−2s+1₋₁

1−ξp(p)^A+μ+1p(−2s+1)(A+μ+1)

= ξp(Πp,1)^−a2ξp(Πp,2)^−a1p^A(s−1)

1−ξp(p)p^−2s

Rp(m, A;s).

Hence we see that I_p^(m)(d(ap);s)

=

⎧⎪

⎨

⎪⎩

ξp(Πp,1)^−a2ξp(Πp,2)^−a1p^A(s−1)(1−p^−2s+1)⁻¹(1−p^−2s) (m= 0), ξp(Πp,1)^−a2ξp(Πp,2)^−a1p^A(s−1)(1−p^−2s)Rp(m, A;s) (m∈p^−AZp),

0 (m∈p^−AZ_p),

where A= ordKap=a1+a2(ap∈Π^a_p,1¹Π^a_p,2²O_K,p^× ) and μ= ordpm.

Letv=∞. The Iwasawa decomposition ofS∞n(x)∞d(a∞) =n(X)d(Y)u∞, where

X =− x

x²+ N(a∞)², Y = a∞

x+ N(a∞)i, u∞ = (x+ N(a∞)i)⁻¹

−x N(a∞)

−N(a∞) −x

implies that

I_∞^(m)(d(a∞);s) = (−1)^κ+2a^s−κ/2−1_∞ a∞s+κ/2+1

×

Re^−2πimx(x+ N(a∞)i)^{−s−κ/2−1}(x−N(a∞)i)^−s+κ/2+1dx.

Fory >0,α, β∈C(Re(α+β)>1) andm∈R, it is known that

Re^−2πimx(x+iy)^−α(x−iy)^−βdx

=

⎧⎪

⎪⎪

⎪⎪

⎪⎨

⎪⎪

⎪⎪

⎪⎪

⎩

i^β−α2^α+βπ^α+βm^α+β−1

e^2πymΓ(α)Γ(β) Φ(4πym;α, β) (m >0), i^β−α2^2−α−βπy^1−α−βΓ(α+β−1)

Γ(α)Γ(β) (m= 0), i^β−α2^α+βπ^α+β|m|^α+β−1

e^2πy|m|Γ(α)Γ(β) Φ(4πy|m|;β, α) (m <0) (cf. [Miy]), where

Φ(z;A, B) = _∞

0

e^−zu(u+ 1)^A−1u^B−1du.

Using the notation of the classical Whittaker functionWν,μ(z), we get Φ(z;A, B) =e^z/2z^−(A+B)/2Γ(B)W(A−B)/2,(A+B−1)/2(z).

Hence we obtain

Re^−2πimx(x+iy)^−α(x−iy)^−βdx

=

⎧⎪

⎪⎪

⎪⎪

⎪⎨

⎪⎪

⎪⎪

⎪⎪

⎩

i^β−απ^(α+β)/2y^−(α+β)/2m^{(α+β−2)/2}Γ(α)⁻¹

×W(α−β)/2,(α+β−1)/2(4πym) (m >0),

i^β−α2^2−α−βπy^1−α−βΓ(α+β−1)Γ(α)⁻¹Γ(β)⁻¹ (m= 0), i^β−απ^(α+β)/2y^−(α+β)/2|m|^{(α+β−2)/2}Γ(β)⁻¹

×W(β−α)/2,(α+β−1)/2(4πy|m|) (m <0).

This equation implies that I_∞^(m)(d(a∞);s)

= i^κ+2a^−κ/2−1_∞ a∞κ/2+1

×

⎧⎪

⎪⎪

⎨

⎪⎪

⎪⎩

π^sm^s−1Γ(s+κ/2 + 1)⁻¹Wκ/2+1,s−1/2(4πN(a∞)m) (m >0), a^1−s_∞ a∞1−s2^2−2sπ Γ(2s−1)

Γ(s+κ/2 + 1)Γ(s−κ/2−1) (m= 0), π^s|m|^s−1Γ(s−κ/2−1)⁻¹W−κ/2−1,s−1/2(4πN(a∞)|m|) (m <0).

Note thatΓ(s+κ/2 + 1) =Pκ/2(s)Γ(s) andΓ(s−κ/2−1) =

κ/2−1

j=−1

(s−κ/2 +j)⁻¹Γ(s) = (−1)^κ/2+1Pκ/2(1−s)⁻¹Γ(s) sinceΓ(X+ 1) =XΓ(X).