betaFeedback

WorksMathematics

On the invariants of L-functions of degree 2, II: twisted conductor

Abstract

We continue our study of the invariants of the LL-functions of deree 2 twisted by Dirichlet characters. In a previous paper, we proved that twisting does not change degree and internal shift. Our aim now is to show a corresponding theorem for the conductor. This paper is intended as the second part of a series aiming to prove a general version of the Weil converse theorem with minimal assumptions on the shape of the functional equation of twists.

Dedicated to Helmut Maier on the occasion of his 70th birthday

Mathematics Subject Classification (2010): 11M41

Introduction

We continue our study of the invariants of the LL-functions of degree 2 twisted by Dirichlet characters. We keep the general notation introduced in [6], where we proved that twisting does not change degree and internal shift. Here we show a corresponding result for the conductor. This paper is intended as the second part of a series aiming to prove a general version of the Weil converse theorem with minimal assumptions on the shape of the functional equation of twists. For the readers’ convenience, we recall all the necessary notation in the next section.

Theorem. Let F∈S♯F \in S^{\sharp} be of degree 2 with integer conductor qFq_F. Let DD be a squarefree positive integer, D≡1(modqF)D \equiv1 \pmod{q_F}, such that FF splits polynomially of degree 2 at every prime p∣Dp \mid D. Moreover, suppose that FF is saturated (mod DD) and for every Dirichlet character χ(modD)\chi\pmod D, the twist Fχ∗F^{\chi^*} belongs to S♯S^{\sharp}. Then the conductors of Fχ∗F^{\chi^*}, FF and χ\chi are related by

qFχ∗=qFfχ2.(1)q_{F^{\chi^*}} = q_F f_\chi^2. \tag*{(1)}

We expect that a suitable version of (1), namely

qFχ=qFfχdF,(2)q_{F^\chi} = q_F f_\chi^{d_F}, \tag*{(2)}

holds in much greater generality. In particular, we expect (2) to hold for every F∈SF \in S, with any degree dFd_F and integer conductor qFq_F, and every primitive character χ(modfχ)\chi\pmod{f_\chi}, (fχ,qF)=1(f_\chi,q_F)=1, such that the twist FχF^\chi also belongs to SS; see conjecture (III.4) in [3]. Note that (2) holds for all classical LL-functions.

Acknowledgements

This research was partially supported by the Istituto Nazionale di Alta Matematica, by grant PRIN 2022 “The arithmetic of motives and LL-functions” and by grant 2021/41/BST1/00241 “Analytic methods in number theory” from the National Science Centre, Poland.

Notation

In this paper

F(s)=∑n=1∞a(n)ns(σ>1)F(s)=\sum_{n=1}^{\infty}\frac{a(n)}{n^s}\qquad(\sigma>1)

stands for a generic function from the extended Selberg class S#\mathcal{S}^{\#}. Recall that every such function satisfies a functional equation of the form

γ(s)F(s)=ωγ(1−s‾)‾F(1−s‾),\gamma(s)F(s)=\omega\overline{\gamma(1-\overline{s})}F(1-\overline{s}),

where

γ(s)=Qs∏j=1rΓ(λjs+μj)\gamma(s)=Q^s\prod_{j=1}^{r}\Gamma(\lambda_j s+\mu_j)

and Q>0Q>0, λj>0\lambda_j>0, ℜ(μj)≥0\Re(\mu_j)\geq0, ∣ω∣=1|\omega|=1. The Selberg class S\mathcal{S} is the subclass of the functions F∈S#F\in\mathcal{S}^{\#} having a rather general Euler product expansion and satisfying the Ramanujan conjecture a(n)≪nεa(n)\ll n^\varepsilon. Degree, conductor, internal shift and root number of FF are defined, respectively, by

dF=2∑j=1rλj,qF=(2πQ)2∏j=1rλj2λj,θF=ℑ(∑j=1rμj),ωF∗=ω∏j=1rλj−2iℑ(μj).d_F=2\sum_{j=1}^{r}\lambda_j,\qquad q_F=(2\pi Q)^2\prod_{j=1}^{r}\lambda_j^{2\lambda_j},\qquad\theta_F=\Im\left(\sum_{j=1}^{r}\mu_j\right),\qquad\omega_F^*=\omega\prod_{j=1}^{r}\lambda_j^{-2i\Im(\mu_j)}.

We refer to our survey papers [2, 1, 8, 9] for definitions, examples, and the basic theory of the Selberg class.

For a Dirichlet character χ\chi and real numbers α\alpha and β\beta, we denote by Fχ(s)F^\chi(s) and F(s,β,α)F(s,\beta,\alpha) the twists

Fχ(s)=∑n=1∞a(n)χ(n)nsandF(s,β,α)=∑n=1∞a(n)nse(−βn−αn).F^\chi(s)=\sum_{n=1}^{\infty}\frac{a(n)\chi(n)}{n^s} \qquad\text{and}\qquad F(s,\beta,\alpha)=\sum_{n=1}^{\infty}\frac{a(n)}{n^s}e(-\beta n-\alpha\sqrt{n}).

If Fχ∈S#F^\chi\in\mathcal{S}^{\#}, then its degree, conductor, internal shift and root number are denoted, respectively, by dFχd_{F^\chi}, qFχq_{F^\chi}, θFχ\theta_{F^\chi} and ωFχ∗\omega_{F^\chi}^*. In the important special case β=0\beta=0 and α≠0\alpha\ne0, F(s,0,α)F(s,0,\alpha) is called the standard twist of FF and is denoted by F(s,α)F(s,\alpha). Concerning the standard twist, we define the spectrum of FF as

Spec⁡(F)={α>0:a(nα)≠0},where nα=qFdFαdF and a(x)=0 if x∉N.\operatorname{Spec}(F)=\{\alpha>0:a(n_\alpha)\ne0\},\quad\text{where }n_\alpha=q_F^{d_F}\alpha^{d_F}\text{ and }a(x)=0\text{ if }x\notin\mathbb{N}.

The relevance of Spec⁡(F)\operatorname{Spec}(F) to the polar structure of F(s,α)F(s,\alpha) is reported in Lemma 3.

We say that F∈S#F\in\mathcal{S}^{\#} splits polynomially of degree 2 at a prime pp if

F(s)=∏j=12(1−αj(p)ps)−1∑p∤na(n)nsF(s)=\prod_{j=1}^{2}\left(1-\frac{\alpha_j(p)}{p^s}\right)^{-1}\sum_{p\nmid n}\frac{a(n)}{n^s}

for certain complex numbers αj(p)≠0\alpha_j(p)\ne0. We say that F∈S#F\in\mathcal{S}^{\#} is saturated (mod DD) if for every integer MM and every residue class aa (mod DD), (a,D)=1(a,D)=1, there exists a positive integer ν≡a\nu\equiv a (mod DD) with (ν,M)=1(\nu,M)=1 such that a(ν)≠0a(\nu)\ne0. Moreover, we say that F∈S#F\in\mathcal{S}^{\#} is normalized if θF=0\theta_F=0 and the first non-zero Dirichlet coefficient of FF equals 1.

For a Dirichlet character χ\chi, we denote by χ∗\chi^* the primitive character inducing χ\chi. As usual, we denote by fχf_\chi the conductor of χ\chi, thus χ∗\chi^* is a Dirichlet character (mod fχf_\chi). Sometimes, in cases where no confusion may occur, when χ\chi is a character (mod qq) we simply write q∗q^* instead of fχf_\chi. Finally, the Gauss sum of χ\chi is denoted by τ(χ)\tau(\chi).

Lemmas

Lemma 1. Let DD be a positive squarefree integer and (a,D)=1(a,D)=1, and suppose that F∈S♯F\in S^{\sharp} with dF=2d_F=2 splits polynomially of degree 2 at all primes p∣Dp\mid D. Then for σ>1\sigma>1 we have

F(s,a/D,α)=1φ(D)∑χ(modD)∑m∣(Dfχ)2Bm(F)f(χ,m,a)msFχ∗(s,mα),(3)F(s,a/D,\alpha)=\frac{1}{\varphi(D)}\sum_{\chi\pmod D}\sum_{m\mid\left(\frac{D}{f_\chi}\right)^2}\frac{B_m(F)f(\chi,m,a)}{m^s}F^{\chi^*}(s,\sqrt{m\alpha}), \tag*{(3)}

where

f(χ,m,a)=μ(Dfχ)τ(χ∗)‾χ∗(Dfχ)χ∗(am)∏p∣mp≠0.f(\chi,m,a)=\mu\left(\frac{D}{f_\chi}\right)\overline{\tau(\chi^*)}\chi^*\left(\frac{D}{f_\chi}\right)\chi^*(am)\prod_{\substack{p\mid m\\p\ne0}}.

Proof. This is a special case of Lemma 3 in [6] with λ=1/2\lambda=1/2. Note that in such a Lemma 3, the analog of equation (3) has m∣(Dfχ)∗m\mid\left(\frac{D}{f_\chi}\right)^* in place of m∣(Dfχ)2m\mid\left(\frac{D}{f_\chi}\right)^2. Here we have simply m∣(Dfχ)2m\mid\left(\frac{D}{f_\chi}\right)^2 thanks to the hypothesis that FF splits polynomially of degree 2 at all primes p∣Dp\mid D. □

Lemma 2. Let F∈S♯F\in S^{\sharp} with dF=2d_F=2 be normalized. Then for β>0\beta>0 and α∈R\alpha\in\mathbb{R} we have

F(s,β,α)=−ωF∗e−iπ2ξFe(α24β)(βqF)2s−1∑n=1∞a(n)nse(nβqF−αβqFn)+h(s)F(s,\beta,\alpha)=-\omega_F^*e^{-\frac{i\pi}{2}\xi_F}e\left(\frac{\alpha^2}{4\beta}\right)(\beta\sqrt{q_F})^{2s-1}\sum_{n=1}^{\infty}\frac{a(n)}{n^s}e\left(\frac{n}{\beta q_F}-\frac{\alpha}{\beta\sqrt{q_F}}\sqrt{n}\right)+h(s)

for σ>1\sigma>1, where h(s)h(s) is holomorphic for σ>1/2\sigma>1/2.

Proof. This is the Lemma in [5] with a slightly changed notation. □

Lemma 3. The standard twist F(s,α)F(s,\alpha) of F∈S♯F\in S^{\sharp} is meromorphic on C\mathbb{C} for every α∈R∖{0}\alpha\in\mathbb{R}\setminus\{0\}. Moreover, if F∈S♯F\in S^{\sharp} has dF=2d_F=2 and θF=0\theta_F=0, then for α∈Spec⁡(F)\alpha\in\operatorname{Spec}(F) we have

res⁡s=3/4F(s,α)=c0(F)a(nα)‾nα1/4,(4)\operatorname{res}_{s=3/4}F(s,\alpha)=c_0(F)\frac{\overline{a(n_\alpha)}}{n_\alpha^{1/4}}, \tag*{(4)}

where c0(F)≠0c_0(F)\ne0 is a constant depending on FF and nα=qFα2/4n_\alpha=q_F\alpha^2/4.

Proof. This follows from Theorems 1 and 2 in [4]. □

We remark that (4) holds also if 0<α∉Spec⁡(F)0<\alpha\notin\operatorname{Spec}(F), in the sense that both sides equal zero in this case.

Lemma 4. Let FF, DD and χ\chi be as in the Theorem. Then dFχ∗=2d_{F^{\chi^*}}=2 and θFχ∗=θF\theta_{F^{\chi^*}}=\theta_F.

Proof. This is the main result of part I of the series, see [6]. □

Lemma 5. If the Theorem is true for every entire F∈S♯F\in S^{\sharp} with dF=2d_F=2, then it holds true for all F∈S♯F\in S^{\sharp} with dF=2d_F=2.

Proof. Let F∈S♯F\in S^{\sharp} with dF=2d_F=2 have a pole of order mF>0m_F>0 at s=1s=1, fix a prime p0∤Dp_0\nmid D and write

G(s)=(1−p0+1p0s+p0p02s)mF.G(s)=\left(1-\frac{p_0+1}{p_0^s}+\frac{p_0}{p_0^{2s}}\right)^{m_F}.

We easily check that G∈S♯G\in S^{\sharp}, and has degree 0 and conductor qG=p02mFq_G=p_0^{2m_F}. Thus the function F1(s)=G(s)F(s)F_1(s)=G(s)F(s) belongs to S♯S^{\sharp}, is entire and has dF1=2d_{F_1}=2 and qF1=qGqF=p02mFqFq_{F_1}=q_Gq_F=p_0^{2m_F}q_F. Moreover, it is easily seen that Gχ∈S♯G^\chi\in S^{\sharp} for every character χ\chi, and qGχ=qGq_{G^\chi}=q_G. Hence applying the Theorem to F1F_1 we get

qF1χ∗=qF1fχ2.q_{F_1^{\chi^*}}=q_{F_1}f_\chi^2.
endsmid=1{ends_mid=1}

But F1χ=GχFχF_1^\chi=G^\chi F_\chi and qF1χ∗=qGχ∗qFχ∗=p02mFqFχ∗q_{F_1^{\chi^*}}=q_{G_{\chi^*}}q_{F_{\chi^*}}=p_0^{2m_F}q_{F_{\chi^*}}. Therefore

p02mFqFχ∗=p02mFqFχfχ2p_0^{2m_F}q_{F\chi^*}=p_0^{2m_F}q_{F\chi}f_\chi^2

and the result follows.

Lemma 6. Let ψ(modq)\psi\pmod q be a Dirichlet character induced by ψ∗(modq∗)\psi^*\pmod{q^*}. Then

τ(ψ)=μ(q/q∗)ψ∗(q/q∗)τ(ψ∗).\tau(\psi)=\mu(q/q^*)\psi^*(q/q^*)\tau(\psi^*).

Proof. This is a well-known property of the Gauss sums, see Theorem 9.10 in [7].

Proof of the Theorem

Using Lemma 5, we may assume without loss of generality that FF is entire. Shifting the variable ss by a suitable purely imaginary constant and multiplying by an appropriate non-zero constant, we can further assume that FF is normalized, thus we may apply Lemma 2. Taking into account that D≡1(modqF)D\equiv1\pmod{q_F}, for σ>1\sigma>1 we obtain

F(s,1/D,α)=ea1+b1se(α2D4)∑n=1∞a(n)‾nse(nDqF−αDqFn)+h(s)=ea1+b1se(α2D4)∑n=1∞a(n)‾nse(nqF−αDqFn)+h(s)=ea1+b1se(α2D4)∑n=1∞a(n)nse(nDqF−αDqFn)+h(s),\begin{aligned} F(s,1/D,\alpha)&=e^{a_1+b_1s}e\left(\frac{\alpha^2D}{4}\right)\sum_{n=1}^{\infty}\frac{\overline{a(n)}}{n^s}e\left(\frac{nD}{q_F}-\frac{\alpha D}{\sqrt{q_F}}\sqrt{n}\right)+h(s)\\ &=e^{a_1+b_1s}e\left(\frac{\alpha^2D}{4}\right)\sum_{n=1}^{\infty}\frac{\overline{a(n)}}{n^s}e\left(\frac{n}{q_F}-\frac{\alpha D}{\sqrt{q_F}}\sqrt{n}\right)+h(s)\\ &=e^{a_1+b_1s}e\left(\frac{\alpha^2D}{4}\right)\sum_{n=1}^{\infty}\frac{a(n)}{n^s}e\left(\frac{nD}{q_F}-\frac{\alpha D}{\sqrt{q_F}}\sqrt{n}\right)+h(s), \end{aligned}

where h(s)h(s) is holomorphic for σ>1/2\sigma>1/2. Applying Lemma 2 once more we see that the last expression equals

ea2+b2se(α2D4(1−D))∑n=1∞a(n)‾nse(αDn)+h(s)=ea2+b2se(α2D4(1−D))F(s,αD)+h(s).(5)\begin{aligned} e^{a_2+b_2s}e\left(\frac{\alpha^2D}{4}(1-D)\right)\sum_{n=1}^{\infty}\frac{\overline{a(n)}}{n^s}e(\alpha D\sqrt{n})+h(s)\\ =e^{a_2+b_2s}e\left(\frac{\alpha^2D}{4}(1-D)\right)F(s,\alpha D)+h(s). \tag*{(5)} \end{aligned}

Next, applying Lemma 1 with a=1a=1, we see that for σ>1\sigma>1

F(s,1/D,α)=1φ(D)∑χ(modD)∑m∣(Dfχ)2Bm(F)f(χ,m,1)msFχ∗(s,mα).(6)F(s,1/D,\alpha)=\frac{1}{\varphi(D)}\sum_{\chi\pmod D}\sum_{m\mid\left(\frac{D}{f_\chi}\right)^2}\frac{B_m(F)f(\chi,m,1)}{m^s}F^{\chi^*}(s,\sqrt{m}\alpha). \tag*{(6)}

Comparing (6) and (5), we arrive at the following identity involving various standard twists

ea2+b2se(α2D4(1−D))F(s,αD)+h(s)=1φ(D)∑χ(modD)∑m∣(Dfχ)2Bm(F)f(χ,m,1)msFχ∗(s,mα).(7)\begin{aligned} e^{a_2+b_2s}e\left(\frac{\alpha^2D}{4}(1-D)\right)F(s,\alpha D)+h(s)\\ =\frac{1}{\varphi(D)}\sum_{\chi\pmod D}\sum_{m\mid\left(\frac{D}{f_\chi}\right)^2}\frac{B_m(F)f(\chi,m,1)}{m^s}F^{\chi^*}(s,\sqrt{m}\alpha). \tag*{(7)} \end{aligned}

Note that (7) is proved for σ>1\sigma>1, but both sides are meromorphic for σ>1/2\sigma>1/2 thanks to Lemmas 2 and 3; thus (7) holds in this larger half-plane. Observe that θFχ∗=0\theta_{F\chi^*}=0 by Lemma 4 and hence, according to Lemma 3, Fχ∗(s,mα)F^{\chi^*}(s,\sqrt{m\alpha}) has a pole at s=3/4s=3/4 whenever mα∈Spec⁡(Fχ∗)\sqrt{m\alpha}\in\operatorname{Spec}(F^{\chi^*}). Taking residues of both sides of (7) and using Lemma 3 again, we obtain

A0(F)e(α2D4(1−D))a(qFα2D2/4)‾(qFα2D2/4)1/4=1φ(D)∑χ(modD)∑m∣(D/fχ)2Bm(F)f(χ,m,1)m3/4c0(Fχ∗)×a(qχmα2/4)χ(qχmα2/4)‾(qχmα2/4)1/4.(8)\begin{aligned} A_0(F)e\left(\frac{\alpha^2D}{4}(1-D)\right)\frac{\overline{a(q_F\alpha^2D^2/4)}}{(q_F\alpha^2D^2/4)^{1/4}} ={}&\frac{1}{\varphi(D)}\sum_{\chi\pmod D}\sum_{m\mid(D/f_\chi)^2}\frac{B_m(F)f(\chi,m,1)}{m^{3/4}}c_0(F^{\chi^*})\\ &\times\frac{\overline{a(q_\chi m\alpha^2/4)\chi(q_\chi m\alpha^2/4)}}{(q_\chi m\alpha^2/4)^{1/4}}. \tag*{(8)} \end{aligned}

where A0(F)≠0A_0(F)\ne0 and, for simplicity, we write qχq_\chi instead of qFχ∗q_{F\chi^*}.

Let

k(χ,m):=qχqFmD2wherem∣(Dfχ)2,k(\chi,m):=\frac{q_\chi}{q_F}\frac{m}{D^2}\qquad\text{where}\qquad m\mid\left(\frac{D}{f_\chi}\right)^2,
k(χ):=k(χ,(D/fχ)2),k0:=max⁡χ(modD)k(χ).k(\chi):=k\left(\chi,(D/f_\chi)^2\right),\qquad k_0:=\max_{\chi\pmod D}k(\chi).

Now in (8) we choose

α=2Dνk0qF(ν∈N),\alpha=\frac{2}{D}\sqrt{\frac{\nu}{k_0q_F}}\qquad(\nu\in\mathbb{N}),

hence

α24D(1−D)=νk0D1−DqF,qF4α2D2=νk0,qχm4α2=qχqFmD=k(χ,m)k0.\frac{\alpha^2}{4}D(1-D)=\frac{\nu}{k_0D}\frac{1-D}{q_F},\qquad\frac{q_F}{4}\alpha^2D^2=\frac{\nu}{k_0},\qquad\frac{q_\chi m}{4}\alpha^2=\frac{q_\chi}{q_F}\frac{m}{D}=\frac{k(\chi,m)}{k_0}.

Thus (8) implies that

e(νk0D1−DqF)a(νk0)‾=∑χ(modD)∑m∣(D/fχ)2cF(χ,m)‾a(k(χ,m)νk0)χ(ν)‾.(9)e\left(\frac{\nu}{k_0D}\frac{1-D}{q_F}\right)\overline{a\left(\frac{\nu}{k_0}\right)} =\sum_{\chi\pmod D}\sum_{m\mid(D/f_\chi)^2}\overline{c_F(\chi,m)}a\left(\frac{k(\chi,m)\nu}{k_0}\right)\overline{\chi(\nu)}. \tag*{(9)}

with certain complex coefficients cF(χ,m)c_F(\chi,m) such that

cF(χ,(D/fχ)2)≠0.(10)c_F\left(\chi,(D/f_\chi)^2\right)\ne0. \tag*{(10)}

Next we show that k0=1k_0=1. We observe that k0≥k(χ0,D2)=1k_0\ge k(\chi_0,D^2)=1 since fχ0=1f_{\chi_0}=1. Suppose, by contradiction, that k0>1k_0>1. If k0∈Qk_0\in\mathbb{Q} then we write

k0=ab(a,b)=1, a>1.k_0=\frac{a}{b}\qquad(a,b)=1,\ a>1.

Let

a∗:={aif k0∈Q,1if k0∉Q.a^*:=\begin{cases} a & \text{if } k_0\in\mathbb{Q},\\ 1 & \text{if } k_0\notin\mathbb{Q}. \end{cases}

Therefore, we have that ν/k0∉N\nu/k_0\notin\mathbb{N} if (ν,a∗)=1(\nu,a^*)=1, and hence a(ν/k0)=0a(\nu/k_0)=0. Recalling (9), for all (ν,a∗)=1(\nu,a^*)=1 we have that

∑χ(modD)∑m∣(D/fχ)2k(χ,m)/k0∈QcF(χ,m)‾a(k(χ,m)νk0)χ(ν)‾=0.(11)\sum_{\chi\pmod D}\sum_{\substack{m\mid(D/f_\chi)^2\\ k(\chi,m)/k_0\in\mathbb{Q}}}\overline{c_F(\chi,m)}a\left(\frac{k(\chi,m)\nu}{k_0}\right)\overline{\chi(\nu)}=0. \tag*{(11)}

When the quotient k(χ,m)/k0k(\chi,m)/k_0 is rational we write

k(χ,m)k0=A(χ,m)B(χ,m)\frac{k(\chi,m)}{k_0}=\frac{A(\chi,m)}{B(\chi,m)}

with A(χ,m),B(χ,m)∈NA(\chi,m), B(\chi,m) \in\mathbb{N} and (A(χ,m),B(χ,m))=1(A(\chi,m),B(\chi,m))=1. If k(χ,m)<k0k(\chi,m)<k_0 then B(χ,m)>1B(\chi,m)>1. Let now

M:=∏χ (mod D)k(χ,m)/k0∈Q,k(χ,m)<k0∏m∣(D/fχ)2B(χ,m).M:=\prod_{\substack{\chi\ (\mathrm{mod}\ D)\\ k(\chi,m)/k_0\in\mathbb{Q},\,k(\chi,m)<k_0}}\prod_{m\mid(D/f_\chi)^2}B(\chi,m).

For (ν,M)=1(\nu,M)=1 we have

k(χ,m)k0ν=A(χ,m)νB(χ,m)∉N,\frac{k(\chi,m)}{k_0}\nu=\frac{A(\chi,m)\nu}{B(\chi,m)}\notin\mathbb{N},

and hence for such ν\nu we see that

a(k(χ,m)k0ν)=0.a\left(\frac{k(\chi,m)}{k_0}\nu\right)=0.

Moreover, for k(χ,m)=k0k(\chi,m)=k_0 we have m=(D/fχ)2m=(D/f_\chi)^2. Thus, for (ν,M)=1(\nu,M)=1, (11) reads as

a(ν)∑χ (mod D)k(χ)=k0cF(χ)χ(ν)=0.a(\nu)\sum_{\substack{\chi\ (\mathrm{mod}\ D)\\ k(\chi)=k_0}}c_F(\chi)\chi(\nu)=0.

where cF(χ):=c(χ,(D/fχ)2)c_F(\chi):=c(\chi,(D/f_\chi)^2). Note that, according to (10), these coefficients are non-zero. Since FF is saturated (mod DD), for every residue class a(modD)a\pmod D, (a,D)=1(a,D)=1, we can find ν≡a(modD)\nu\equiv a\pmod D, (ν,M)=1(\nu,M)=1, such that a(ν)≠0a(\nu)\ne0. Hence for every (a,D)=1(a,D)=1 we have

∑χ (mod D)k(χ)=k0cF(χ)χ(a)=0.\sum_{\substack{\chi\ (\mathrm{mod}\ D)\\ k(\chi)=k_0}}c_F(\chi)\chi(a)=0.

This contradicts the linear independence of characters, thus we have k0=1k_0=1.

Finally we show that k(χ)=1k(\chi)=1 for every character χ(modD)\chi\pmod D. Since k0=1k_0=1, (9) reads as

e(νD1−DqF)a(ν)‾=∑χ (mod D)∑m∣(D/fχ)2k(χ,m)∈QcF(χ,m)a(k(χ,m)ν)χ(ν).(12)e\left(\frac{\nu}{D}\frac{1-D}{q_F}\right)\overline{a(\nu)} = \sum_{\chi\ (\mathrm{mod}\ D)} \sum_{\substack{m\mid(D/f_\chi)^2\\ k(\chi,m)\in\mathbb{Q}}} c_F(\chi,m)a\bigl(k(\chi,m)\nu\bigr)\chi(\nu). \tag*{(12)}

For (ν,M)=1(\nu,M)=1, we have k(χ,m)ν∉Nk(\chi,m)\nu\notin\mathbb{N}, and hence the corresponding terms on the right hand side vanish. In particular, non-zero terms must have k(χ,m)=1k(\chi,m)=1. Recalling that in the latter case m=(D/fχ)2m=(D/f_\chi)^2, equation (12) takes the form

e(νD1−DqF)a(ν)‾=a(ν)‾∑χ (mod D)k(χ)=1cF(χ)χ(ν)‾.e\left(\frac{\nu}{D}\frac{1-D}{q_F}\right)\overline{a(\nu)} = \overline{a(\nu)} \sum_{\substack{\chi\ (\mathrm{mod}\ D)\\ k(\chi)=1}}c_F(\chi)\overline{\chi(\nu)}.

As before, we use the assumption that FF is saturated (mod DD), thus getting that

e(aD1−DqF)=∑χ (mod D)k(χ)=1cF(χ)χ(a)‾(13)e\left(\frac{a}{D}\frac{1-D}{q_F}\right) = \sum_{\substack{\chi\ (\mathrm{mod}\ D)\\ k(\chi)=1}}c_F(\chi)\overline{\chi(a)} \tag*{(13)}

for all (a,D)=1(a,D)=1. Suppose that the sum on the right hand side of (13) does not contain all characters (mod DD). For instance, suppose that k(ψ)≠1k(\psi)\ne1 for a certain ψ(modD)\psi\pmod D. Multiplying both sides of (13) by ψ(a)\psi(a) and summing over all (a,D)=1(a,D)=1 we obtain, by the orthogonality of characters, that

∑a (mod D)(a,D)=1ψ(a)e(aD1−DqF)=0.\sum_{\substack{a\ (\mathrm{mod}\ D)\\ (a,D)=1}}\psi(a)e\left(\frac{a}{D}\frac{1-D}{q_F}\right)=0.

Recalling Lemma 6, we see that this sum equals

ψ(1−DqF)τ(ψ)=ψ(1−DqF)μ(D/fχ)ψ∗(D/fχ)τ(ψ∗)\psi\left(\frac{1-D}{q_F}\right)\tau(\psi)=\psi\left(\frac{1-D}{q_F}\right)\mu(D/f_\chi)\psi^*(D/f_\chi)\tau(\psi^*)

and hence is non-zero since DD is square-free. This contradiction shows that k(χ)=1k(\chi)=1 and consequently qχ=qFfχ2q_\chi=q_F f_\chi^2 for all χ\chi (mod DD). The proof is complete.

Jerzy Kaczorowski, Faculty of Mathematics and Computer Science, A. Mickiewicz University, 61-614 Poznań, Poland. e-mail: [email protected]

Alberto Perelli, Dipartimento di Matematica, Università di Genova, via Dodecaneso 35, 16146 Genova, Italy. e-mail: [email protected]

References

  1. [1]J.Kaczorowski - Axiomatic theory of L-functions: the Selberg class - In Analytic Number Theory, C.I.M.E. Summer School, Cetraro (Italy) 2002, ed. by A.Perelli and C.Viola, 133–209, Springer L.N. 1891, 2006.DOI
  2. [2]J.Kaczorowski, A.Perelli - The Selberg class: a survey - In Number Theory in Progress, Proc. Conf. in Honor of A.Schinzel, ed. by K.Györy et al., 953–992, de Gruyter 1999.DOI
  3. [3]J.Kaczorowski, A.Perelli - On the structure of the Selberg class, II: invariants and conjectures - J. reine angew. Math. 524 (2000), 73–96.DOI
  4. [4]J.Kaczorowski, A.Perelli - On the structure of the Selberg class, VI: non-linear twists - Acta Arith. 116 (2005), 315–341.DOI
  5. [5]J.Kaczorowski, A.Perelli - A weak converse theorem for degree 2 L-functions with conductor 1 - Publ. Res. Inst. Math. Sci. 53 (2017), no. 2, 337–347.DOI
  6. [6]J.Kaczorowski, A.Perelli - On the invariants of L-functions of degree 2, I: twisted degree and internal shift - Preprint.
  7. [7]H.L.Montgomery, R. C.Vaughan - Multiplicative Number Theory. I. Classical Theory - Cambridge U. P. 2007.
  8. [8]A.Perelli - A survey of the Selberg class of L-functions, part I - Milan J. Math. 73 (2005), 19–52.DOI
  9. [9]A.Perelli - A survey of the Selberg class of L-functions, part II - Riv. Mat. Univ. Parma (7) 3* (2004), 83–118.

Paper details

vlnc:W_au7jg5ga
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents