input
stringclasses
1 value
output
stringlengths
0
3.63k
instruction
stringlengths
0
3.63k
data_source
stringclasses
1 value
\begin{defi} Soit $C_c^{\infty}(F^*)$ l'espace des applications $F^* \longrightarrow \mathbb{C}$ localement constantes à support compact. \end{defi}
\begin{defi} Soit $C_c^{\infty}(F^*)$ l'espace des applications $F^* \longrightarrow \mathbb{C}$ localement constantes à support compact. \end{defi}
\begin{defi} On définit les fonctions zêta locales par : $$ \mathcal{Z}(f, \chi) = \int_{F^*} f(x) \chi(x) \, dx^{\times} $$ où $f \in C_c^{\infty}(F)$ et $\chi$ est un caractère multiplicatif de $F^*$, égal à $\eta |.|^s$. \end{defi}
\begin{defi} On définit les fonctions zêta locales par : $$ \mathcal{Z}(f, \chi) = \int_{F^*} f(x) \chi(x) \, dx^{\times} $$ où $f \in C_c^{\infty}(F)$ et $\chi$ est un caractère multiplicatif de $F^*$, égal à $\eta |.|^s$. \end{defi}
\section{Fonctions L-abéliennes et théorème de Tate} \subsection{Dualité pour les groupes abéliens localement compacts}
\section{Fonctions L-abéliennes et théorème de Tate} \subsection{Dualité pour les groupes abéliens localement compacts}
\begin{proof}~ Cela résulte immédiatement de la décomposition topologique $F^{*} \simeq \OO^{\times} \times \mathbb{Z}$. \end{proof}
\begin{proof}~ Cela résulte immédiatement de la décomposition topologique $F^{*} \simeq \OO^{\times} \times \mathbb{Z}$. \end{proof}
\subsection{Théorie de Fourier sur les corps $p$-adiques} On se donne $p$ premier et $F$ une extension finie de $\mathbb{Q}_{p}$. On note $\OO$ son anneau de valuation et $\varpi \in \OO$ une uniformisante. $q$ est le cardinal du corps résiduel.
\subsection{Théorie de Fourier sur les corps $p$-adiques} On se donne $p$ premier et $F$ une extension finie de $\mathbb{Q}_{p}$. On note $\OO$ son anneau de valuation et $\varpi \in \OO$ une uniformisante. $q$ est le cardinal du corps résiduel.
\begin{proof}~ Il suffit de prouver le corollaire pour $\psi_0 : x \mapsto \psi_p ( \operatorname{Tr}_{F/\Q_p}(x)$ car alors pour chaque $\psi \neq \1$, il existe $a \in \mathbb{Q}_{p}^*$ tel que $\psi=\left(\psi_0\right)_a$. \begin{center} \begin{tikzpicture} \def\a{1.5} \def\b{2} \path (-\a,0) node (A) {$F$} (\a,0) node (B) {$\hat{F}$} (0,-\b) node[align=center] (C) {$F$}; \begin{scope}[nodes={midway,scale=1}] \draw[->] (A)--(B) node[above]{$b \mapsto \psi_b$}; \draw[->] (A)--(C.120) node[left]{$b \mapsto a b$}; \draw[->] (C.60)--(B) node[right]{$a \mapsto\left(\psi_0\right)_a $}; \end{scope} \end{tikzpicture} \end{center} Où $(\psi_b = (\psi_0 )_{ab} )$. \\ On prend donc $\psi=\psi_0$ comme référence. $\theta$ est un isomorphisme bijectif, il faut vérifier sa bicontinuité. \\
\begin{proof}~ Il suffit de prouver le corollaire pour $\psi_0 : x \mapsto \psi_p ( \operatorname{Tr}_{F/\Q_p}(x)$ car alors pour chaque $\psi \neq \1$, il existe $a \in \mathbb{Q}_{p}^*$ tel que $\psi=\left(\psi_0\right)_a$. \begin{center} \begin{tikzpicture} \def\a{1.5} \def\b{2} \path (-\a,0) node (A) {$F$} (\a,0) node (B) {$\hat{F}$} (0,-\b) node[align=center] (C) {$F$}; \begin{scope}[nodes={midway,scale=1}] \draw[->] (A)--(B) node[above]{$b \mapsto \psi_b$}; \draw[->] (A)--(C.120) node[left]{$b \mapsto a b$}; \draw[->] (C.60)--(B) node[right]{$a \mapsto\left(\psi_0\right)_a $}; \end{scope} \end{tikzpicture} \end{center} Où $(\psi_b = (\psi_0 )_{ab} )$. \\ On prend donc $\psi=\psi_0$ comme référence. $\theta$ est un isomorphisme bijectif, il faut vérifier sa bicontinuité. \\
\begin{proof}~ \underline{Séparation} : Soient $\chi, \phi \in \widehat{G}$ avec $\chi \neq \phi$. Il existe $g \in G$ tel que $\chi(g) \neq \phi(g)$. $\exists U \ni \chi(g) \; , \; V \ni \phi(g)$ ouverts tq $U \cap V = \varnothing$. Alors, $W(\{g\}, U) \ni \chi$ et $W(\{g\}, V) \ni \phi$ sont deux ouverts de $\hat{G}$ disjoints. \underline{Inversion continue} : $$ \mathcal{I}: \left\{\begin{array}{l} \hat{G} \rightarrow \hat{G} \\ \chi \mapsto \chi^{-1} = \bar{\chi} \end{array} \quad \mathcal{I}^{-1}(W(K, U)) = W(K, \bar{U}) \text{ , c'est un ouvert.}$ \underline{Produit continu} : $$ \mu: \left\{\begin{array}{l} \hat{G} \times \hat{G} \rightarrow \hat{G} \\ (\psi, \phi) \mapsto \psi \phi \end{array} \quad \text{Par le lemme précédent il suffit de montrer que } \mu^{-1} \left(B_K(\chi, \varepsilon)\right) \text{est ouvert.}$ Soit $(\psi, \phi) \in \mu^{-1}\left(B_K(\chi, \varepsilon)\right)$. \\ On a $\psi \phi \in B_K(\chi, \varepsilon)$ d'où $m = \max\limits_{k \in K}|(\psi \phi)(k) - \chi(k)| < \varepsilon$. \\ Pour $\eta = \frac{\varepsilon - m}{4}$, on a que $\forall \tilde{\psi} \in B_K(\Psi, \eta) \; , \; \forall \tilde{\phi} \in B_K(\phi, \eta) \; , \; \forall k \in K$ : $$|(\widetilde{\psi} \widetilde{\phi})(k) - \chi(k)| \leqslant |(\widetilde{\psi}(k) - \psi(k)) \widetilde{\phi}(k)| + |\psi(k)(\Phi(k) - \phi(k))| + |(\psi \phi)(k) - \chi(k)| < \eta + \eta + m \leqslant \frac{\varepsilon - m}{2} + m < \varepsilon. $$ Ce qui montre la continuité de $\mu$. \end{proof}
\begin{proof}~ \underline{Séparation} : Soient $\chi, \phi \in \widehat{G}$ avec $\chi \neq \phi$. Il existe $g \in G$ tel que $\chi(g) \neq \phi(g)$. $\exists U \ni \chi(g) \; , \; V \ni \phi(g)$ ouverts tq $U \cap V = \varnothing$. Alors, $W(\{g\}, U) \ni \chi$ et $W(\{g\}, V) \ni \phi$ sont deux ouverts de $\hat{G}$ disjoints. \underline{Inversion continue} : $$ \mathcal{I}: \left\{\begin{array}{l} \hat{G} \rightarrow \hat{G} \\ \chi \mapsto \chi^{-1} = \bar{\chi} \end{array} \quad \mathcal{I}^{-1}(W(K, U)) = W(K, \bar{U}) \text{ , c'est un ouvert.}$ \underline{Produit continu} : $$ \mu: \left\{\begin{array}{l} \hat{G} \times \hat{G} \rightarrow \hat{G} \\ (\psi, \phi) \mapsto \psi \phi \end{array} \quad \text{Par le lemme précédent il suffit de montrer que } \mu^{-1} \left(B_K(\chi, \varepsilon)\right) \text{est ouvert.}$ Soit $(\psi, \phi) \in \mu^{-1}\left(B_K(\chi, \varepsilon)\right)$. \\ On a $\psi \phi \in B_K(\chi, \varepsilon)$ d'où $m = \max\limits_{k \in K}|(\psi \phi)(k) - \chi(k)| < \varepsilon$. \\ Pour $\eta = \frac{\varepsilon - m}{4}$, on a que $\forall \tilde{\psi} \in B_K(\Psi, \eta) \; , \; \forall \tilde{\phi} \in B_K(\phi, \eta) \; , \; \forall k \in K$ : $$|(\widetilde{\psi} \widetilde{\phi})(k) - \chi(k)| \leqslant |(\widetilde{\psi}(k) - \psi(k)) \widetilde{\phi}(k)| + |\psi(k)(\Phi(k) - \phi(k))| + |(\psi \phi)(k) - \chi(k)| < \eta + \eta + m \leqslant \frac{\varepsilon - m}{2} + m < \varepsilon. $$ Ce qui montre la continuité de $\mu$. \end{proof}
\subsection{Fonctions zêtas locales (p-adiques)} \begin{defi} Un caractère multiplicatif de $F^{*}$ est un morphisme continu $\chi:(F^{*}, \times) \rightarrow(\mathbb{C}^*, \times)$. \end{defi}
\subsection{Fonctions zêtas locales (p-adiques)} \begin{defi} Un caractère multiplicatif de $F^{*}$ est un morphisme continu $\chi:(F^{*}, \times) \rightarrow(\mathbb{C}^*, \times)$. \end{defi}
\begin{lemme} Pour toutes $f, g \in C_{c}^{\infty}(F)$ et tout caractère $\chi = \eta |.|^s$ avec $0 < \Re(s) < 1$, on a : $$ \mathcal{Z}\left(\hat{f}, \chi^{\vee}\right) \mathcal{Z}(g, \chi) = \mathcal{Z}(f, \chi) \mathcal{Z}\left(\hat{g}, \chi^{\vee}\right) $$ où chaque fonction $\mathcal{Z}$ est sur son domaine de convergence. L'identité se prolonge méromorphiquement à $\mathbb{C}$. \end{lemme} \begin{proof} Pour $0 < \Re(s) < 1$, le membre de gauche est dominé par : $$ \begin{aligned} & \int_{F^*} \hat{f}(x) \chi^{\vee}(x) dx^{\times} \int_{F^*} g(y) \chi(y) dy^{\times} \quad \text { (tout converge absolument) } \\ & = \int_{(F^*)^2} \hat{f}(x) g(y) |x| \chi\left(y x^{-1}\right) dx^{\times} dy^{\times} \\ & = \int_{\left(F^*\right)^2} \hat{f}(x) g\left(y^{\prime} x\right) |x| \chi\left(y^{\prime}\right) dx^{\times} dy^{\prime}^{\times} \\ & = \int_{F^*} \chi(y) \left(\int_{F^*} \hat{f}(x) g(x y) |x| dx^{\times}\right) dy^{\times}\\ & \end{aligned} $$ L'intégrale intérieure vaut $\left(1-q^{-1}\right)^{-1} \int_F \hat{f}(x) g(x y) d x$. $$ \begin{aligned} &\stackrel{Plancherel}{=} \left(1-q^{-1}\right)^{-1} \int_F f(x) |y|^{-1} \hat{g}\left(y^{-1} x\right) d x \\ & \; \; \; \stackrel{x^{\prime} = y^{\prime} x }{=} \left(1-q^{-1}\right)^{-1} \int_F f\left(y x^{\prime}\right) \hat{g}\left(x^{\prime}\right) d x x\end{aligned} $$ On obtient finalement une expression symétrique en $f$ et $g$. \end{proof}
\begin{lemme} Pour toutes $f, g \in C_{c}^{\infty}(F)$ et tout caractère $\chi = \eta |.|^s$ avec $0 < \Re(s) < 1$, on a : $$ \mathcal{Z}\left(\hat{f}, \chi^{\vee}\right) \mathcal{Z}(g, \chi) = \mathcal{Z}(f, \chi) \mathcal{Z}\left(\hat{g}, \chi^{\vee}\right) $$ où chaque fonction $\mathcal{Z}$ est sur son domaine de convergence. L'identité se prolonge méromorphiquement à $\mathbb{C}$. \end{lemme} \begin{proof} Pour $0 < \Re(s) < 1$, le membre de gauche est dominé par : $$ \begin{aligned} & \int_{F^*} \hat{f}(x) \chi^{\vee}(x) dx^{\times} \int_{F^*} g(y) \chi(y) dy^{\times} \quad \text { (tout converge absolument) } \\ & = \int_{(F^*)^2} \hat{f}(x) g(y) |x| \chi\left(y x^{-1}\right) dx^{\times} dy^{\times} \\ & = \int_{\left(F^*\right)^2} \hat{f}(x) g\left(y^{\prime} x\right) |x| \chi\left(y^{\prime}\right) dx^{\times} dy^{\prime}^{\times} \\ & = \int_{F^*} \chi(y) \left(\int_{F^*} \hat{f}(x) g(x y) |x| dx^{\times}\right) dy^{\times}\\ & \end{aligned} $$ L'intégrale intérieure vaut $\left(1-q^{-1}\right)^{-1} \int_F \hat{f}(x) g(x y) d x$. $$ \begin{aligned} &\stackrel{Plancherel}{=} \left(1-q^{-1}\right)^{-1} \int_F f(x) |y|^{-1} \hat{g}\left(y^{-1} x\right) d x \\ & \; \; \; \stackrel{x^{\prime} = y^{\prime} x }{=} \left(1-q^{-1}\right)^{-1} \int_F f\left(y x^{\prime}\right) \hat{g}\left(x^{\prime}\right) d x x\end{aligned} $$ On obtient finalement une expression symétrique en $f$ et $g$. \end{proof}
\begin{proof}~ \\ L'équation fonctionnelle vient de celle entre $\mathcal{Z}$ et $\gamma$. $\frac{\mathcal{Z}(f(x))}{L(\chi)}$ est holomorphe pour $\operatorname{Re}(s)>0$. $\frac{\mathcal{Z}\left(\hat{f}, \chi^\vee\right)}{L\left(\chi^\vee\right)}$ est holomorphe pour $\operatorname{Re}(1-s)>0$ ie $\operatorname{Re}(s)<1$. $\varepsilon$ est holomorphe. \end{proof} Pour aller plus loin : $$ f=f_1+f(0) \1_{\OO} \quad \text { avec } f_1 \in C_c^{\infty}\left(F^*\right). $$ $\mathcal{Z}\left(f_1, \chi\right)$ est un polynôme en $q^{ \pm s}$ et $\frac{1}{L(\chi)}=\left\{\begin{array}{lc}1 &\text { si } \chi \text { est ramifié } \\ 1-q^{-s} &\text {sinon} \end{array}\right.$. et $\mathcal{Z}\left(\1_{\OO}, \chi\right)=\left\{\begin{array}{lc} 0 &\text { si } \chi \text { est ramifié } \\ \frac{\operatorname{Vol}(\OO^\times,dx^\times)}{1-q^{-s}} = \operatorname{Vol}(\OO^\times,dx^\times) L(\chi) &\text {sinon} \end{array} \right.$.
\begin{proof}~ \\ L'équation fonctionnelle vient de celle entre $\mathcal{Z}$ et $\gamma$. $\frac{\mathcal{Z}(f(x))}{L(\chi)}$ est holomorphe pour $\operatorname{Re}(s)>0$. $\frac{\mathcal{Z}\left(\hat{f}, \chi^\vee\right)}{L\left(\chi^\vee\right)}$ est holomorphe pour $\operatorname{Re}(1-s)>0$ ie $\operatorname{Re}(s)<1$. $\varepsilon$ est holomorphe. \end{proof} Pour aller plus loin : $$ f=f_1+f(0) \1_{\OO} \quad \text { avec } f_1 \in C_c^{\infty}\left(F^*\right). $$ $\mathcal{Z}\left(f_1, \chi\right)$ est un polynôme en $q^{ \pm s}$ et $\frac{1}{L(\chi)}=\left\{\begin{array}{lc}1 &\text { si } \chi \text { est ramifié } \\ 1-q^{-s} &\text {sinon} \end{array}\right.$. et $\mathcal{Z}\left(\1_{\OO}, \chi\right)=\left\{\begin{array}{lc} 0 &\text { si } \chi \text { est ramifié } \\ \frac{\operatorname{Vol}(\OO^\times,dx^\times)}{1-q^{-s}} = \operatorname{Vol}(\OO^\times,dx^\times) L(\chi) &\text {sinon} \end{array} \right.$.
\begin{lemme} On a $\varepsilon(\chi, \psi) \varepsilon\left(\chi^\vee, \psi\right)=\chi(-1)$. \end{lemme} \begin{proof} $\mathcal{Z}(\hat{\hat{f}}, \chi) \stackrel{\text { Fourier }}{=} \int_{F^*} f(-x) \chi(x) d x^\times=\chi(-1) \mathcal{Z}(f, \chi)$. \\ Or $\frac{\mathcal{Z}\hat{\hat{f}}, \chi)}{L(\chi)}=\varepsilon\left(\chi^\vee, \psi\right) \frac{\mathcal{Z}\left(\hat{f}, \chi^\vee\right)}{L\left(\chi^\vee\right)}=\varepsilon\left(\chi^\vee, \psi\right) \varepsilon(\chi, \psi) \frac{\mathcal{Z}(f, \chi)}{L(\chi)}$. \end{proof}
\begin{lemme} On a $\varepsilon(\chi, \psi) \varepsilon\left(\chi^\vee, \psi\right)=\chi(-1)$. \end{lemme} \begin{proof} $\mathcal{Z}(\hat{\hat{f}}, \chi) \stackrel{\text { Fourier }}{=} \int_{F^*} f(-x) \chi(x) d x^\times=\chi(-1) \mathcal{Z}(f, \chi)$. \\ Or $\frac{\mathcal{Z}\hat{\hat{f}}, \chi)}{L(\chi)}=\varepsilon\left(\chi^\vee, \psi\right) \frac{\mathcal{Z}\left(\hat{f}, \chi^\vee\right)}{L\left(\chi^\vee\right)}=\varepsilon\left(\chi^\vee, \psi\right) \varepsilon(\chi, \psi) \frac{\mathcal{Z}(f, \chi)}{L(\chi)}$. \end{proof}
\begin{lemme}~ \begin{enumerate} \item[(i)] Pour $\Re(s) > 0$, l'intégrale converge absolument. \\ $s \mapsto \mathcal{Z}\left(f, \eta |.|^s\right)$ est holomorphe sur le demi-plan $\Re(s) > 0$. \item[(ii)] $s \mapsto \mathcal{Z}\left(f, \eta |.|^s \right)$ admet un prolongement méromorphe à $\mathbb{C}$. \end{enumerate} \end{lemme}
\begin{lemme}~ \begin{enumerate} \item[(i)] Pour $\Re(s) > 0$, l'intégrale converge absolument. \\ $s \mapsto \mathcal{Z}\left(f, \eta |.|^s\right)$ est holomorphe sur le demi-plan $\Re(s) > 0$. \item[(ii)] $s \mapsto \mathcal{Z}\left(f, \eta |.|^s \right)$ admet un prolongement méromorphe à $\mathbb{C}$. \end{enumerate} \end{lemme}
\begin{proof}(\cref{lem2})~ Par définition du conducteur, on a : $\left.\psi\right|_{\varpi^m \OO} \equiv 1, \left.\psi\right|_{\varpi^{m-1} \OO} \not \equiv 1 \text {. }$ \\ Donc, si $n \geqslant m$, on a $\varpi^n \OO \subseteq \varpi^m \OO$ et l'intégrale vaut $\operatorname{Vol}(\varpi^m \OO, dx)$. \\ Supposons que $n < m$, alors $\exists a \in \varpi^{n} \OO$ tel que $\psi(a) \neq 1$. Alors, $$ \int_{\varpi^n \OO} \psi(x) \,dx = \int_F \1_{\varpi^n \OO}(x) \psi(x) \,dx $$ $$ \stackrel{dx \text{ de Haar}}{=} \int_F \1_{\varpi^n \OO}(x+a) \psi(x+a) \,dx $$ $$ \stackrel{\text{car }a \in \varpi^n \OO}{=} \psi(a) \int_F \1_{\varpi^n \OO}(x) \psi(x) \,dx $$ $$ = \psi(a) \int_{\varpi^n \OO} \psi(x) \,dx $$ Ainsi, $\overbrace{(1-\psi(a))}^{\neq 0} \int_{\varpi^n \OO} \psi(x) \,dx = 0$ et donc $\int_{\varpi^n \OO} \psi(x) \,dx = 0$. \end{proof}
\begin{proof}(\cref{lem2})~ Par définition du conducteur, on a : $\left.\psi\right|_{\varpi^m \OO} \equiv 1, \left.\psi\right|_{\varpi^{m-1} \OO} \not \equiv 1 \text {. }$ \\ Donc, si $n \geqslant m$, on a $\varpi^n \OO \subseteq \varpi^m \OO$ et l'intégrale vaut $\operatorname{Vol}(\varpi^m \OO, dx)$. \\ Supposons que $n < m$, alors $\exists a \in \varpi^{n} \OO$ tel que $\psi(a) \neq 1$. Alors, $$ \int_{\varpi^n \OO} \psi(x) \,dx = \int_F \1_{\varpi^n \OO}(x) \psi(x) \,dx $$ $$ \stackrel{dx \text{ de Haar}}{=} \int_F \1_{\varpi^n \OO}(x+a) \psi(x+a) \,dx $$ $$ \stackrel{\text{car }a \in \varpi^n \OO}{=} \psi(a) \int_F \1_{\varpi^n \OO}(x) \psi(x) \,dx $$ $$ = \psi(a) \int_{\varpi^n \OO} \psi(x) \,dx $$ Ainsi, $\overbrace{(1-\psi(a))}^{\neq 0} \int_{\varpi^n \OO} \psi(x) \,dx = 0$ et donc $\int_{\varpi^n \OO} \psi(x) \,dx = 0$. \end{proof}
\subsection{Calculs auxiliaires} On cherche à obtenir des formules pour $\frac{\mathcal{Z}\left(\hat{g}, \chi^{\vee}\right)}{\mathcal{Z}(g, \chi)}$.\\ Dans ce passage, $\psi$ est un caractère additif non trivial de $F$ avec $n_\psi:=\min \left\{n \in \mathbb{Z}\,\mid \, \psi|_{\varpi^n \OO} \equiv 1\right\}$. \\ $\chi$ quant à lui, est un caractère multiplicatif unitaire. \\ On note $U_0=\OO^{\times}$, $U_n=1+\varpi^n \OO$ pour $n \geqslant 1$ avec $n_\chi:=\min \left\{n \in \mathbb{N} \;|\; \chi|_{U_n} \equiv 1\right\}$.\\ On note enfin $G(\chi, \psi)=\int_{\OO^{\times}} \chi(x) \psi(x) \, dx^{\times}$.
\subsection{Calculs auxiliaires} On cherche à obtenir des formules pour $\frac{\mathcal{Z}\left(\hat{g}, \chi^{\vee}\right)}{\mathcal{Z}(g, \chi)}$.\\ Dans ce passage, $\psi$ est un caractère additif non trivial de $F$ avec $n_\psi:=\min \left\{n \in \mathbb{Z}\,\mid \, \psi|_{\varpi^n \OO} \equiv 1\right\}$. \\ $\chi$ quant à lui, est un caractère multiplicatif unitaire. \\ On note $U_0=\OO^{\times}$, $U_n=1+\varpi^n \OO$ pour $n \geqslant 1$ avec $n_\chi:=\min \left\{n \in \mathbb{N} \;|\; \chi|_{U_n} \equiv 1\right\}$.\\ On note enfin $G(\chi, \psi)=\int_{\OO^{\times}} \chi(x) \psi(x) \, dx^{\times}$.
\begin{rmk}~ \begin{enumerate} \item Si $\chi=\chi^\vee$ on a $1-s=s$ c'est-à-dire $s=1/2$ et $\eta=\bar{\eta}$, c'est-à-dire $\eta$ est réel. \\ Ainsi $\chi(-1)= \pm 1$ et $\varepsilon(\chi, \psi)^2=\chi(-1)$ donc $\varepsilon(\chi, \psi) \in \mathbb{U}_4$. \item Plus généralement: $$ \begin{aligned} &\varepsilon\left(\eta|\cdot|^{1 / 2}, \psi\right) \varepsilon \left(\bar{\eta}|\cdot|^{1 / 2}, \psi\right)=\eta(-1), \\ & \text { Et } \overline{\mathcal{Z}(f, \chi)}=\mathcal{Z}(\bar{f}, \bar{\chi}) , \, \overline{\mathcal{Z}\left(\hat{f}, \chi^\vee\right)}=\mathcal{Z}\left(\bar{\hat{f}}, \overline{\chi^\vee}\right), \, \overline{\hat{f}}(y)=\hat{\overline{f}}(-y), \\ & \text { Et } \mathcal{Z}\left(\hat{\overline{f}}, \bar{\chi}^\vee\right) \overline{\chi}^\vee(-1)=\gamma(\bar{\chi}, \psi) \mathcal{Z}(\bar{f}, \bar{\chi}) \bar{\chi}^\vee(-1). \end{aligned} $$ D'où $\overline{\gamma(\chi, \psi)}=\gamma(\bar{\chi}, \psi) \bar{\chi}^\vee(-1)$. \\ Ainsi $\overline{\varepsilon(\eta \mid \cdot \mid^{1 / 2}, \psi)}=\varepsilon(\bar{\eta}\mid \cdot \mid^{1 / 2}, \psi) \eta(-1)$, et on retrouve donc $\left|\varepsilon\left(\eta|\cdot|^{-1 / 2}, \psi\right)\right|=1$. \item Hormis les énoncés qui utilisent l'inversion de Fourier, les résultats tiennent $\forall \psi \not \equiv 1$. Si on veut être plus précis, on note le facteur epsilon $\varepsilon(\chi, \psi, dx)$. \\ On a: $ \begin{aligned}[t] & \rightarrow \forall t>0 \quad \varepsilon(\chi, \psi, t d x)=t \varepsilon(\chi, \psi, d x) \\ & \rightarrow \forall a \in F^* \quad \varepsilon\left(\chi, \psi_a, d x\right)=\chi(a)|a|^{-1} \varepsilon(\chi, \psi, d x) \\ & \rightarrow \varepsilon\left(\chi |\cdot|^{s}, \psi, d x\right)=|\varpi|^{(-n_\psi+n_\chi) s} \varepsilon(\chi, \psi, d x). \end{aligned} $ \end{enumerate} \end{rmk}
\begin{rmk}~ \begin{enumerate} \item Si $\chi=\chi^\vee$ on a $1-s=s$ c'est-à-dire $s=1/2$ et $\eta=\bar{\eta}$, c'est-à-dire $\eta$ est réel. \\ Ainsi $\chi(-1)= \pm 1$ et $\varepsilon(\chi, \psi)^2=\chi(-1)$ donc $\varepsilon(\chi, \psi) \in \mathbb{U}_4$. \item Plus généralement: $$ \begin{aligned} &\varepsilon\left(\eta|\cdot|^{1 / 2}, \psi\right) \varepsilon \left(\bar{\eta}|\cdot|^{1 / 2}, \psi\right)=\eta(-1), \\ & \text { Et } \overline{\mathcal{Z}(f, \chi)}=\mathcal{Z}(\bar{f}, \bar{\chi}) , \, \overline{\mathcal{Z}\left(\hat{f}, \chi^\vee\right)}=\mathcal{Z}\left(\bar{\hat{f}}, \overline{\chi^\vee}\right), \, \overline{\hat{f}}(y)=\hat{\overline{f}}(-y), \\ & \text { Et } \mathcal{Z}\left(\hat{\overline{f}}, \bar{\chi}^\vee\right) \overline{\chi}^\vee(-1)=\gamma(\bar{\chi}, \psi) \mathcal{Z}(\bar{f}, \bar{\chi}) \bar{\chi}^\vee(-1). \end{aligned} $$ D'où $\overline{\gamma(\chi, \psi)}=\gamma(\bar{\chi}, \psi) \bar{\chi}^\vee(-1)$. \\ Ainsi $\overline{\varepsilon(\eta \mid \cdot \mid^{1 / 2}, \psi)}=\varepsilon(\bar{\eta}\mid \cdot \mid^{1 / 2}, \psi) \eta(-1)$, et on retrouve donc $\left|\varepsilon\left(\eta|\cdot|^{-1 / 2}, \psi\right)\right|=1$. \item Hormis les énoncés qui utilisent l'inversion de Fourier, les résultats tiennent $\forall \psi \not \equiv 1$. Si on veut être plus précis, on note le facteur epsilon $\varepsilon(\chi, \psi, dx)$. \\ On a: $ \begin{aligned}[t] & \rightarrow \forall t>0 \quad \varepsilon(\chi, \psi, t d x)=t \varepsilon(\chi, \psi, d x) \\ & \rightarrow \forall a \in F^* \quad \varepsilon\left(\chi, \psi_a, d x\right)=\chi(a)|a|^{-1} \varepsilon(\chi, \psi, d x) \\ & \rightarrow \varepsilon\left(\chi |\cdot|^{s}, \psi, d x\right)=|\varpi|^{(-n_\psi+n_\chi) s} \varepsilon(\chi, \psi, d x). \end{aligned} $ \end{enumerate} \end{rmk}
\begin{proof}~ \begin{enumerate} \item[(i)] $ \Rightarrow$ Clair. $\Leftarrow$ Montrons que pour $m \geqslant 1$, $\chi^{-1}(N(\frac{1}{m}))$ est un voisinage de $e$. \\ Par hypothèse, $\exists U \ni e$ ouvert de $G$ tel que $\chi(U) \subseteq N(1)$. Par continuité de $\left\{\begin{array}{l}G^m \rightarrow G \\ (g_1, \ldots, g_m) \mapsto g_1 \dots g_m\end{array}\right.$ , $ \exists V \ni e$ ouvert tel que $V^{(m)} \subseteq U$. \\ Ainsi $\chi(V^{(m)}) \subseteq N(1)$ et par le Lemme, $\chi(V) \subseteq N(\frac{1}{m})$, d'où $V \subseteq \chi^{-1}(N(\frac{1}{m}))$. \\ Soit $B \subseteq \U$ un ouvert. Montrons que $\chi^{-1}(B)$ est ouvert :\\ Soit $g \in \chi^{-1}(B)$. On a $\chi(g) \in B$. $\exists m \geqslant 1$ assez grand tel que $\chi(g) N(\frac{1}{m}) \subseteq B$. Soit $V \ni e$ un voisinage de $G$ tel que $ \chi(V) \subseteq N(\frac{1}{m}) $ alors : $ \chi(g V) = \chi(g) \chi(V) \subseteq \chi(g) N(\frac{1}{m}) \subseteq B$. \\ Donc $gV \subseteq \chi^{-1}(B)$. Ainsi, $\chi^{-1}(B)$ est un voisinage de chacun de ses points, c'est un ouvert. \\ \item[(ii)] On sait que les $B_K(\1, \varepsilon)=W(K, B(1, \varepsilon))$ forment une base de voisinages de $\1$, et alors les $W(K, N(1/m))$ aussi, (pour $K \subseteq G$ compact, et $m \geqslant 1$). \\ $W(K, N(1))$ étant un voisinage ouvert de $\1$, il suffit de voir que $\forall K \subseteq G$ compact, $\forall m \geqslant 1$, $\exists L \subseteq G$ compact tel que $\mathcal{W}(L, N(1)) \subseteq W(K, N(\frac{1}{m}))$. \\ Soient $K \subseteq G$ compact et $m \geqslant 1$. On pose $L:=(K \cup\{1\})^{(m)}$, c'est un compact de $G$. Pour $\chi \in \mathcal{W}(L, N(1))$, on a : $\chi(L) \subseteq N(1)$, d'où $\chi((K \cup\{1\})^{(m)}) \subseteq N(1)$, donc $\chi(K \cup\{1\}) \subseteq N(\frac{1}{m})$. Ainsi, $\chi \in W(K, N(\frac{1}{m}))$, ce qui conclut. \end{enumerate} \end{proof}
\begin{proof}~ \begin{enumerate} \item[(i)] $ \Rightarrow$ Clair. $\Leftarrow$ Montrons que pour $m \geqslant 1$, $\chi^{-1}(N(\frac{1}{m}))$ est un voisinage de $e$. \\ Par hypothèse, $\exists U \ni e$ ouvert de $G$ tel que $\chi(U) \subseteq N(1)$. Par continuité de $\left\{\begin{array}{l}G^m \rightarrow G \\ (g_1, \ldots, g_m) \mapsto g_1 \dots g_m\end{array}\right.$ , $ \exists V \ni e$ ouvert tel que $V^{(m)} \subseteq U$. \\ Ainsi $\chi(V^{(m)}) \subseteq N(1)$ et par le Lemme, $\chi(V) \subseteq N(\frac{1}{m})$, d'où $V \subseteq \chi^{-1}(N(\frac{1}{m}))$. \\ Soit $B \subseteq \U$ un ouvert. Montrons que $\chi^{-1}(B)$ est ouvert :\\ Soit $g \in \chi^{-1}(B)$. On a $\chi(g) \in B$. $\exists m \geqslant 1$ assez grand tel que $\chi(g) N(\frac{1}{m}) \subseteq B$. Soit $V \ni e$ un voisinage de $G$ tel que $ \chi(V) \subseteq N(\frac{1}{m}) $ alors : $ \chi(g V) = \chi(g) \chi(V) \subseteq \chi(g) N(\frac{1}{m}) \subseteq B$. \\ Donc $gV \subseteq \chi^{-1}(B)$. Ainsi, $\chi^{-1}(B)$ est un voisinage de chacun de ses points, c'est un ouvert. \\ \item[(ii)] On sait que les $B_K(\1, \varepsilon)=W(K, B(1, \varepsilon))$ forment une base de voisinages de $\1$, et alors les $W(K, N(1/m))$ aussi, (pour $K \subseteq G$ compact, et $m \geqslant 1$). \\ $W(K, N(1))$ étant un voisinage ouvert de $\1$, il suffit de voir que $\forall K \subseteq G$ compact, $\forall m \geqslant 1$, $\exists L \subseteq G$ compact tel que $\mathcal{W}(L, N(1)) \subseteq W(K, N(\frac{1}{m}))$. \\ Soient $K \subseteq G$ compact et $m \geqslant 1$. On pose $L:=(K \cup\{1\})^{(m)}$, c'est un compact de $G$. Pour $\chi \in \mathcal{W}(L, N(1))$, on a : $\chi(L) \subseteq N(1)$, d'où $\chi((K \cup\{1\})^{(m)}) \subseteq N(1)$, donc $\chi(K \cup\{1\}) \subseteq N(\frac{1}{m})$. Ainsi, $\chi \in W(K, N(\frac{1}{m}))$, ce qui conclut. \end{enumerate} \end{proof}
\subsubsection*{Construction des caractères additifs de $\mathbb{Q}_p$}
\subsubsection*{Construction des caractères additifs de $\mathbb{Q}_p$}
\begin{defi} Si $\eta$ est un caractère non trivial de $\OO^{\times}$, on définit son conducteur comme le plus grand idéal $\varpi^n \OO$, $n \geq 1$, tel que $\left.\eta\right|_{1+\varpi^n \OO } \equiv 1$. \end{defi}
\begin{defi} Si $\eta$ est un caractère non trivial de $\OO^{\times}$, on définit son conducteur comme le plus grand idéal $\varpi^n \OO$, $n \geq 1$, tel que $\left.\eta\right|_{1+\varpi^n \OO } \equiv 1$. \end{defi}
\begin{proof}~ \begin{enumerate} \item[(i)] Soit $f = \underbrace{\left(f - f(0) \1_{\OO}\right)}_{=: g \in C_c^{\infty}(F^*)} + f(0) \1_{\OO}$. \\ Ainsi, $\mathcal{Z}(g, \chi)$ est défini par une intégrale sur un compact : $$ \int_{F^*} |g(x) \chi(x)|_{\infty} \, dx^{\times} = \int_{F^*} |g(x)|_{\infty} |x|^{\Re(s)} \, dx^{\times} \quad (\eta \text{ est à valeur dans } \U_{\infty}) $$ On voit que cette intégrale converge absolument, et même uniformément convergente pour $\Re(s)$ dans un compact de $\mathbb{R}$. Ainsi, l'intégrale qui définit $\mathcal{Z}\left(g, \eta |.|^s \right)$ a un sens pour tout $s \in \mathbb{C}$ et définit une fonction holomorphe en $s$. \item[(ii)] On sait que $\eta$, donc $\chi$, est trivial sur $1+\varpi^m \OO$ pour un $m \geqslant 1$. \\ Il suffit de traiter le cas de $g = \1_{a\left(1+\varpi^n \OO \right)}$, $n \geqslant m$ : $$ \begin{aligned} \mathcal{Z}(g, \chi) & = \int_{F^*} \1_{a\left(1+\varpi^n \OO\right)}(x) \chi(x) \, dx^{\times} \\ & = \int_{F^*} \1_{a\left(1+\varpi^n \OO \right)}(ax) \chi(ax) \, dx^{\times} \\ & = \chi(a) \int_{F^*} \1_{1+\varpi^n \OO}(x) \chi(x) \, dx^{\times} \\ & = \chi(a) \operatorname{Vol}\left(1+\varpi^n \OO, dx^{\times}\right) \\ & = \eta(a) |a|^s \operatorname{Vol}\left(1+\varpi^m \OO, dx^{\times}\right) . \end{aligned} $$ \end{enumerate} Il reste à traiter le cas $f=\1_{\OO}$ : $$ \begin{aligned} \int_{F^{\times}}|f|_{\infty}|\chi|_{\infty} & =\int_{F^{\times} \cap \OO }|x|^{\sigma} \, dx^{\times} \quad(\sigma=\operatorname{Re}(s)) \\ & =\sum_{n=0}^{+\infty} \int_{\varpi^n \OO^{\times}}|x|^{\sigma} \, dx^{\times} \\ & =\sum_{n=0}^{+\infty} q^{-n \sigma} \operatorname{Vol}\left(\varpi^n \OO^{\times}, dx^{\times}\right) \\ & =\operatorname{Vol}\left(\OO^{\times}, dx^{\times}\right) \sum_{n=0}^{+\infty} q^{-n \sigma} \end{aligned} $$ Soit $\sigma_0>0$. L'intégrale $\mathcal{Z}(f, \chi)$ converge absolument et uniformément sur $\Re(s) \geqslant \sigma_0$, elle définit donc une fonction holomorphe sur $\Re(s)>0$. Soit $\chi=\eta |.|^s$ avec $\Re(s)>0$. $$ \begin{aligned} \mathcal{Z}(f, \chi)=\int_{F^{\times}} \1_{\OO}(x) \chi(x) \, dx^{\times} & =\sum_{n=0}^{+\infty} \int_{\varpi^n \OO^{\times}} \chi(x) \, dx^{\times} \\ & =\left(\sum_{n=0}^{+\infty} \underbrace{\chi(\varpi)^n}_{q^{-n \sigma}}\right) \int_{\OO^{\times}} \underbrace{\chi(x)}_{= \eta (x)} \, dx^{\times}. \end{aligned} $$ ~- Supposons $\eta$ non ramifié, c'est-à-dire $\exists a \in \OO^{\times} , \; \chi(a)=\eta(a) \neq 1$. $$ \begin{aligned} & \int_{\OO^{\times}} \chi(x) \, dx^{\times}=\int_{\OO^{\times}} \chi(a x) \, dx^{\times}=\chi(a) \int_{\OO^{\times}} \chi(x) \, dx^{\times} . \\ & \text{D'où } (1-\chi(a)) \int_{\OO^{\times}} \chi=0 \text{ et } \int_{\OO^{\times}} \chi=0 \text{, donc } \mathcal{Z}(f, \chi)=0 \end{aligned} $$ Ce qui est prolongeable holomorphiquement à tout $\C$. \\ ~- Supposons $\eta$ ramifié et $\left.\chi\right|_{\OO^{\times}} \equiv 1$. $$ \text{Alors } \mathcal{Z}(f, \chi)=\operatorname{Vol}\left(\OO^{\times}, dx^{\times}\right) \frac{1}{1-\chi(\varpi)}=\operatorname{Vol}\left(\OO^{\times}, dx^{\times}\right) \frac{1}{1-q^{-s}} $$ Ceci donne un prolongement méromorphe (pôles en $s \in \frac{2 i \pi \mathbb{Z}}{\log q}$). \end{proof}
\begin{proof}~ \begin{enumerate} \item[(i)] Soit $f = \underbrace{\left(f - f(0) \1_{\OO}\right)}_{=: g \in C_c^{\infty}(F^*)} + f(0) \1_{\OO}$. \\ Ainsi, $\mathcal{Z}(g, \chi)$ est défini par une intégrale sur un compact : $$ \int_{F^*} |g(x) \chi(x)|_{\infty} \, dx^{\times} = \int_{F^*} |g(x)|_{\infty} |x|^{\Re(s)} \, dx^{\times} \quad (\eta \text{ est à valeur dans } \U_{\infty}) $$ On voit que cette intégrale converge absolument, et même uniformément convergente pour $\Re(s)$ dans un compact de $\mathbb{R}$. Ainsi, l'intégrale qui définit $\mathcal{Z}\left(g, \eta |.|^s \right)$ a un sens pour tout $s \in \mathbb{C}$ et définit une fonction holomorphe en $s$. \item[(ii)] On sait que $\eta$, donc $\chi$, est trivial sur $1+\varpi^m \OO$ pour un $m \geqslant 1$. \\ Il suffit de traiter le cas de $g = \1_{a\left(1+\varpi^n \OO \right)}$, $n \geqslant m$ : $$ \begin{aligned} \mathcal{Z}(g, \chi) & = \int_{F^*} \1_{a\left(1+\varpi^n \OO\right)}(x) \chi(x) \, dx^{\times} \\ & = \int_{F^*} \1_{a\left(1+\varpi^n \OO \right)}(ax) \chi(ax) \, dx^{\times} \\ & = \chi(a) \int_{F^*} \1_{1+\varpi^n \OO}(x) \chi(x) \, dx^{\times} \\ & = \chi(a) \operatorname{Vol}\left(1+\varpi^n \OO, dx^{\times}\right) \\ & = \eta(a) |a|^s \operatorname{Vol}\left(1+\varpi^m \OO, dx^{\times}\right) . \end{aligned} $$ \end{enumerate} Il reste à traiter le cas $f=\1_{\OO}$ : $$ \begin{aligned} \int_{F^{\times}}|f|_{\infty}|\chi|_{\infty} & =\int_{F^{\times} \cap \OO }|x|^{\sigma} \, dx^{\times} \quad(\sigma=\operatorname{Re}(s)) \\ & =\sum_{n=0}^{+\infty} \int_{\varpi^n \OO^{\times}}|x|^{\sigma} \, dx^{\times} \\ & =\sum_{n=0}^{+\infty} q^{-n \sigma} \operatorname{Vol}\left(\varpi^n \OO^{\times}, dx^{\times}\right) \\ & =\operatorname{Vol}\left(\OO^{\times}, dx^{\times}\right) \sum_{n=0}^{+\infty} q^{-n \sigma} \end{aligned} $$ Soit $\sigma_0>0$. L'intégrale $\mathcal{Z}(f, \chi)$ converge absolument et uniformément sur $\Re(s) \geqslant \sigma_0$, elle définit donc une fonction holomorphe sur $\Re(s)>0$. Soit $\chi=\eta |.|^s$ avec $\Re(s)>0$. $$ \begin{aligned} \mathcal{Z}(f, \chi)=\int_{F^{\times}} \1_{\OO}(x) \chi(x) \, dx^{\times} & =\sum_{n=0}^{+\infty} \int_{\varpi^n \OO^{\times}} \chi(x) \, dx^{\times} \\ & =\left(\sum_{n=0}^{+\infty} \underbrace{\chi(\varpi)^n}_{q^{-n \sigma}}\right) \int_{\OO^{\times}} \underbrace{\chi(x)}_{= \eta (x)} \, dx^{\times}. \end{aligned} $$ ~- Supposons $\eta$ non ramifié, c'est-à-dire $\exists a \in \OO^{\times} , \; \chi(a)=\eta(a) \neq 1$. $$ \begin{aligned} & \int_{\OO^{\times}} \chi(x) \, dx^{\times}=\int_{\OO^{\times}} \chi(a x) \, dx^{\times}=\chi(a) \int_{\OO^{\times}} \chi(x) \, dx^{\times} . \\ & \text{D'où } (1-\chi(a)) \int_{\OO^{\times}} \chi=0 \text{ et } \int_{\OO^{\times}} \chi=0 \text{, donc } \mathcal{Z}(f, \chi)=0 \end{aligned} $$ Ce qui est prolongeable holomorphiquement à tout $\C$. \\ ~- Supposons $\eta$ ramifié et $\left.\chi\right|_{\OO^{\times}} \equiv 1$. $$ \text{Alors } \mathcal{Z}(f, \chi)=\operatorname{Vol}\left(\OO^{\times}, dx^{\times}\right) \frac{1}{1-\chi(\varpi)}=\operatorname{Vol}\left(\OO^{\times}, dx^{\times}\right) \frac{1}{1-q^{-s}} $$ Ceci donne un prolongement méromorphe (pôles en $s \in \frac{2 i \pi \mathbb{Z}}{\log q}$). \end{proof}
\begin{proof}~ \begin{enumerate} \item[(i)] En considérant $B(1, \varepsilon)$ dans $\mathbb{C}^*$, on observe que $\psi^{-1}(B(1, \varepsilon))$ est un ouvert de $F$ qui contient $0$. \\ Il existe donc $n \in \mathbb{Z}$ tel que $\varpi^n \OO \subseteq \psi^{-1}(B(1, \varepsilon))$, d'où $\psi(\varpi^n \OO) \subseteq B(1, \varepsilon)$.\\ Or $\varpi^n \OO$ est un idéal, donc $\psi(\varpi^n \OO)$ est un sous-groupe de $\C^*$, et il est compact). \\ Ainsi $\psi(\varpi^n \OO) \subseteq \underbrace{\U \cap B(1, \varepsilon) \subseteq N(1)}_{\text{pour }\varepsilon \text{ petit}}$ d'où $\psi(\varpi^n \OO) = \{ 1 \}$ (\cref{lem}). Ainsi $\varpi^n \OO \subseteq \operatorname{Ker} \psi$ et alors $\operatorname{Ker} \psi = \bigcup\limits_{x \in \operatorname{Ker} \psi} x+\varpi^n \OO$ est ouvert. \item[(ii)] Soit $n \in \mathbb{Z}$ tel que $\varpi^n \OO \subseteq \operatorname{Ker} \psi$. On a $F = \bigcup\limits_{N \in \mathbb{Z}} \varpi^N \OO$ (union décroissante). \\ Pour tout $x \in F$, il existe $N \leqslant n$ tel que $x \in \varpi^N \OO$ et $\varpi^n \OO \subseteq \varpi^N \OO$. \\ La restriction de $\psi$ à $ \varpi^N \OO$ se factorise par le groupe fini $\varpi^N \OO / \varpi^n \OO$. Donc $\psi(\varpi^N \OO)$ est un sous-groupe fini de $\C^*$, donc inclus dans $\U_{\infty}$. \end{enumerate} \end{proof}
\begin{proof}~ \begin{enumerate} \item[(i)] En considérant $B(1, \varepsilon)$ dans $\mathbb{C}^*$, on observe que $\psi^{-1}(B(1, \varepsilon))$ est un ouvert de $F$ qui contient $0$. \\ Il existe donc $n \in \mathbb{Z}$ tel que $\varpi^n \OO \subseteq \psi^{-1}(B(1, \varepsilon))$, d'où $\psi(\varpi^n \OO) \subseteq B(1, \varepsilon)$.\\ Or $\varpi^n \OO$ est un idéal, donc $\psi(\varpi^n \OO)$ est un sous-groupe de $\C^*$, et il est compact). \\ Ainsi $\psi(\varpi^n \OO) \subseteq \underbrace{\U \cap B(1, \varepsilon) \subseteq N(1)}_{\text{pour }\varepsilon \text{ petit}}$ d'où $\psi(\varpi^n \OO) = \{ 1 \}$ (\cref{lem}). Ainsi $\varpi^n \OO \subseteq \operatorname{Ker} \psi$ et alors $\operatorname{Ker} \psi = \bigcup\limits_{x \in \operatorname{Ker} \psi} x+\varpi^n \OO$ est ouvert. \item[(ii)] Soit $n \in \mathbb{Z}$ tel que $\varpi^n \OO \subseteq \operatorname{Ker} \psi$. On a $F = \bigcup\limits_{N \in \mathbb{Z}} \varpi^N \OO$ (union décroissante). \\ Pour tout $x \in F$, il existe $N \leqslant n$ tel que $x \in \varpi^N \OO$ et $\varpi^n \OO \subseteq \varpi^N \OO$. \\ La restriction de $\psi$ à $ \varpi^N \OO$ se factorise par le groupe fini $\varpi^N \OO / \varpi^n \OO$. Donc $\psi(\varpi^N \OO)$ est un sous-groupe fini de $\C^*$, donc inclus dans $\U_{\infty}$. \end{enumerate} \end{proof}
\begin{defi} Le bidual de $G$ est le dual $\widehat{\hat{G}}$ de $\hat{G}$. \end{defi} Pour chaque $g \in G$, on dispose d'un morphisme de groupes $ev_{g}: \widehat{G} \longrightarrow \mathbb{U}$ défini par $\chi \mapsto \chi(g)$. \\ Pour voir que $ev_{g}$ est continue, il suffit de montrer que $\mathrm{ev_{g}}^{-1}(N(1))$ est un voisinage de $\1$ (\cref{lem}). \\ Or $\mathrm{ev_{g}}^{-1}(N(1)) = \{\chi \in \widehat{G} \mid \chi(g) \in N(1)\} = \overbrace{W(\{g\}, N(1))}^{\text{ouvert de } \hat{G} \text{ contenant } \1} $. Ainsi, pour tout $g \in G$, $ev_{g} \in \widehat{\hat{G}}$.
\begin{defi} Le bidual de $G$ est le dual $\widehat{\hat{G}}$ de $\hat{G}$. \end{defi} Pour chaque $g \in G$, on dispose d'un morphisme de groupes $ev_{g}: \widehat{G} \longrightarrow \mathbb{U}$ défini par $\chi \mapsto \chi(g)$. \\ Pour voir que $ev_{g}$ est continue, il suffit de montrer que $\mathrm{ev_{g}}^{-1}(N(1))$ est un voisinage de $\1$ (\cref{lem}). \\ Or $\mathrm{ev_{g}}^{-1}(N(1)) = \{\chi \in \widehat{G} \mid \chi(g) \in N(1)\} = \overbrace{W(\{g\}, N(1))}^{\text{ouvert de } \hat{G} \text{ contenant } \1} $. Ainsi, pour tout $g \in G$, $ev_{g} \in \widehat{\hat{G}}$.
\begin{proof}~ \begin{align*} \operatorname{Vol}(\OO^{\times}, dx^{\times}) &= \frac{1}{1 - q^{-1}} \int_{\OO^{\times}} \frac{dx}{|x|} \\ &= \frac{1}{1 - q^{-1}} \int_{\OO^{\times}} dx \\ &= \frac{1}{1 - q^{-1}} \operatorname{Vol}(\OO^{\times}, dx) \\ &= \frac{1}{1 - q^{-1}} \big( \operatorname{Vol}(\OO, dx) - \operatorname{Vol}(\varpi \OO, dx) \big)\\ &= \frac{1}{1 - q^{-1}} \big( \operatorname{Vol}(\OO, dx) - |\varpi|\operatorname{Vol}(\OO, dx) \big)\\ &= \operatorname{Vol}(\OO, dx) \end{align*} \end{proof}
\begin{proof}~ \begin{align*} \operatorname{Vol}(\OO^{\times}, dx^{\times}) &= \frac{1}{1 - q^{-1}} \int_{\OO^{\times}} \frac{dx}{|x|} \\ &= \frac{1}{1 - q^{-1}} \int_{\OO^{\times}} dx \\ &= \frac{1}{1 - q^{-1}} \operatorname{Vol}(\OO^{\times}, dx) \\ &= \frac{1}{1 - q^{-1}} \big( \operatorname{Vol}(\OO, dx) - \operatorname{Vol}(\varpi \OO, dx) \big)\\ &= \frac{1}{1 - q^{-1}} \big( \operatorname{Vol}(\OO, dx) - |\varpi|\operatorname{Vol}(\OO, dx) \big)\\ &= \operatorname{Vol}(\OO, dx) \end{align*} \end{proof}
\begin{lemme} On suppose $n_\chi>0$. Alors: \begin{align*} & \rightarrow \text { Si } n_\psi<n_{\chi}, \quad G(\chi, \psi)=0 \\ & \rightarrow \text{Si } n_\psi=n_{\chi}, \quad|G(\chi, \psi)|_{\infty}^2=\frac{|\OO / \mathcal{D} |^{-1 / 2} }{1-q^{-1}} \operatorname{Vol}\left(U_{n_\psi}, dx^{\times}\right) \\ & \rightarrow \text{Si }n_\psi>n_{\chi}, \quad|G(\chi, \psi)|_{\infty}^2=\frac{|\OO / \mathcal{D} |^{-1 / 2} }{1-q^{-1}}\left(\operatorname{Vol}\left(U_{n_\psi}, dx^{\times}\right)-q^{-1} \operatorname{Vol}\left(U_{n_{\psi^{-1}}}, dx^{\times}\right)\right) \end{align*} \end{lemme}
\begin{lemme} On suppose $n_\chi>0$. Alors: \begin{align*} & \rightarrow \text { Si } n_\psi<n_{\chi}, \quad G(\chi, \psi)=0 \\ & \rightarrow \text{Si } n_\psi=n_{\chi}, \quad|G(\chi, \psi)|_{\infty}^2=\frac{|\OO / \mathcal{D} |^{-1 / 2} }{1-q^{-1}} \operatorname{Vol}\left(U_{n_\psi}, dx^{\times}\right) \\ & \rightarrow \text{Si }n_\psi>n_{\chi}, \quad|G(\chi, \psi)|_{\infty}^2=\frac{|\OO / \mathcal{D} |^{-1 / 2} }{1-q^{-1}}\left(\operatorname{Vol}\left(U_{n_\psi}, dx^{\times}\right)-q^{-1} \operatorname{Vol}\left(U_{n_{\psi^{-1}}}, dx^{\times}\right)\right) \end{align*} \end{lemme}
\begin{defi} Un caractère $\chi$ de $F^*$ est non ramifié si $\left.\chi\right|_{\OO^{\times}} \equiv 1$. Dans la décomposition de la \cref{prop1}, cela revient à dire que $\eta=1$, donc $\chi=|.|^s$. \end{defi}
\begin{defi} Un caractère $\chi$ de $F^*$ est non ramifié si $\left.\chi\right|_{\OO^{\times}} \equiv 1$. Dans la décomposition de la \cref{prop1}, cela revient à dire que $\eta=1$, donc $\chi=|.|^s$. \end{defi}
\begin{proof}~{(Corollaire)} On applique le lemme à $V=G$ où $G$ est un sous-groupe de $\mathbb{C}^*$ inclus dans $N(1)$ et à $\chi=Id$. $G^{(m)} \subseteq G \subseteq N(1)$ d'où $\forall m \, , \, G \subseteq N\left(\frac{1}{m}\right) $ et donc $G \subseteq \bigcap\limits_{m \geqslant 1} N\left(\frac{1}{m}\right) = \{1\}$. \end{proof}
\begin{proof}~{(Corollaire)} On applique le lemme à $V=G$ où $G$ est un sous-groupe de $\mathbb{C}^*$ inclus dans $N(1)$ et à $\chi=Id$. $G^{(m)} \subseteq G \subseteq N(1)$ d'où $\forall m \, , \, G \subseteq N\left(\frac{1}{m}\right) $ et donc $G \subseteq \bigcap\limits_{m \geqslant 1} N\left(\frac{1}{m}\right) = \{1\}$. \end{proof}
\underline{Continuité de $\theta^{-1}$} : \\ Là encore, il suffit de montrer que $\theta(B(0, \varepsilon))$ est un voisinage de $\1$ dans $\hat{F}$. \\ On souhaite donc trouver $K \subseteq F$ compact et $m \geq 1$ tels que $W(K, N(1 / m)) \subseteq \theta(B(0, \varepsilon))$. Soit $x \in F$ tel que $\psi_0(x) \neq 1$. Il existe $m \geq 1$ tel que $\psi_0(x) \notin N(1 / m)$. Posons $M:=\varepsilon^{-1}|x|$ et $K:=\overline{B(0,M)}$. \\ $\begin{array}{lll} \text{Alors, }\theta^{-1}(W(K, N(1 / m))) &= \lbrace y \in F \mid \psi_0 (yK) \subseteq N (1/m) \rbrace &\subseteq \lbrace y \in F \mid x \notin yK \rbrace \\ & &= \lbrace y \in F \mid |x| > |y| M \rbrace \\ & &\subseteq B(O, \underbrace{|x|M^{-1}}_{= \varepsilon}). \end{array}$ \end{proof} Fixons un caractère non trivial de $F$, en l'occurrence $\psi:=\psi_p \circ \operatorname{Tr}_{F / Q_p}$. \\ On note alors $dx$ la mesure de Haar sur $F$ telle que $\operatorname{Vol}(\OO, dx) = | \OO / \mathcal{D} |^{-1/2}$ où $\mathcal{D}^{-1}=\{x \in F \mid \operatorname{Tr}_{F / Q_p}(x y) \in \mathbb{Z}_{p} , \forall y \in \OO\}$. On s'intéresse à l'espace de fonctions (analogue à la classe de Schwartz sur $\mathbb{R}$): $$C_c^{\infty}(F):=\{f: F \rightarrow \mathbb{C} \text{ localement constante et à support compact} \}$$.
\underline{Continuité de $\theta^{-1}$} : \\ Là encore, il suffit de montrer que $\theta(B(0, \varepsilon))$ est un voisinage de $\1$ dans $\hat{F}$. \\ On souhaite donc trouver $K \subseteq F$ compact et $m \geq 1$ tels que $W(K, N(1 / m)) \subseteq \theta(B(0, \varepsilon))$. Soit $x \in F$ tel que $\psi_0(x) \neq 1$. Il existe $m \geq 1$ tel que $\psi_0(x) \notin N(1 / m)$. Posons $M:=\varepsilon^{-1}|x|$ et $K:=\overline{B(0,M)}$. \\ $\begin{array}{lll} \text{Alors, }\theta^{-1}(W(K, N(1 / m))) &= \lbrace y \in F \mid \psi_0 (yK) \subseteq N (1/m) \rbrace &\subseteq \lbrace y \in F \mid x \notin yK \rbrace \\ & &= \lbrace y \in F \mid |x| > |y| M \rbrace \\ & &\subseteq B(O, \underbrace{|x|M^{-1}}_{= \varepsilon}). \end{array}$ \end{proof} Fixons un caractère non trivial de $F$, en l'occurrence $\psi:=\psi_p \circ \operatorname{Tr}_{F / Q_p}$. \\ On note alors $dx$ la mesure de Haar sur $F$ telle que $\operatorname{Vol}(\OO, dx) = | \OO / \mathcal{D} |^{-1/2}$ où $\mathcal{D}^{-1}=\{x \in F \mid \operatorname{Tr}_{F / Q_p}(x y) \in \mathbb{Z}_{p} , \forall y \in \OO\}$. On s'intéresse à l'espace de fonctions (analogue à la classe de Schwartz sur $\mathbb{R}$): $$C_c^{\infty}(F):=\{f: F \rightarrow \mathbb{C} \text{ localement constante et à support compact} \}$$.
\begin{rmk} \begin{enumerate} \item[1)] Si $G$ est discret ses compacts sont les sous-groupes finis. La topologie compacte-ouverte est alors celle de la topologie produit sur $\mathbb{U}^G$. \item[2)] Pour $(\chi_n) \in \hat{G}^{\N}$ et $\chi \in \hat{G}$, $\chi_n \rightarrow \chi \Leftrightarrow \chi_n \text{converge uniformément vers } \chi \text{ sur tout compact de G}$. \item[3)] Si $G$ est $\sigma$-compact (réunion dénombrable de compacts) alors la topologie compacte-ouverte sur $\hat{G}$ est métrisable. \end{enumerate} \end{rmk}
\begin{rmk} \begin{enumerate} \item[1)] Si $G$ est discret ses compacts sont les sous-groupes finis. La topologie compacte-ouverte est alors celle de la topologie produit sur $\mathbb{U}^G$. \item[2)] Pour $(\chi_n) \in \hat{G}^{\N}$ et $\chi \in \hat{G}$, $\chi_n \rightarrow \chi \Leftrightarrow \chi_n \text{converge uniformément vers } \chi \text{ sur tout compact de G}$. \item[3)] Si $G$ est $\sigma$-compact (réunion dénombrable de compacts) alors la topologie compacte-ouverte sur $\hat{G}$ est métrisable. \end{enumerate} \end{rmk}
\begin{rmk} Ici, on a fait jouer un rôle privilégié à $\psi_p$ mais \textit{a posteriori} le lemme vaut en remplaçant $\psi_p$ par n'importe quel morphisme $\psi \neq \1$ et $\Z_p$ par $\operatorname{Cond}(\psi)$. \end{rmk}
\begin{rmk} Ici, on a fait jouer un rôle privilégié à $\psi_p$ mais \textit{a posteriori} le lemme vaut en remplaçant $\psi_p$ par n'importe quel morphisme $\psi \neq \1$ et $\Z_p$ par $\operatorname{Cond}(\psi)$. \end{rmk}
\begin{proof}~ \begin{enumerate} \item[Opt 1 :] $x \mapsto e^{2 i \pi x}$ est un morphisme de $\mathbb{R}$ dans $\mathbb{U}$ trivial sur $\mathbb{Z}$. Par restriction et quotient, cela donne un morphisme $$ \left\{ \begin{aligned} \mathbb{Z}[1 / p] / \mathbb{Z} & \longrightarrow \mathbb{U} \\ x & \longmapsto e^{2 i \pi x} \end{aligned} \quad \text{ de noyau trivial}\right. $$ On en déduit un morphisme $\mathbb{Q}_{p} \rightarrow \mathbb{Q}_{p}/\Z_p \rightarrow \mathbb{U}$ de noyau $\mathbb{Z}_{p}$. \\ On note $\psi_{p} : \mathbb{Q}_{p} \rightarrow \mathbb{U}$, il est continu, non trivial et a pour conducteur $\mathbb{Z}_{p}$. \item[Opt 2 :] Le morphisme est donné par le développement de Hensel. \end{enumerate} \end{proof} \begin{lemme} Tout caractère additif $\psi: \mathbb{Q}_{p} \rightarrow \mathbb{C}^*$ est de la forme $x \mapsto \psi_{p}(a x)$ pour un unique $a \in \mathbb{Q}_{p}$. De plus, $\psi$ est non trivial si et seulement si $a \neq 0$, et dans ce cas, le conducteur de $\psi$ est $a^{-1} \mathbb{Z}_{p}$. \end{lemme}
\begin{proof}~ \begin{enumerate} \item[Opt 1 :] $x \mapsto e^{2 i \pi x}$ est un morphisme de $\mathbb{R}$ dans $\mathbb{U}$ trivial sur $\mathbb{Z}$. Par restriction et quotient, cela donne un morphisme $$ \left\{ \begin{aligned} \mathbb{Z}[1 / p] / \mathbb{Z} & \longrightarrow \mathbb{U} \\ x & \longmapsto e^{2 i \pi x} \end{aligned} \quad \text{ de noyau trivial}\right. $$ On en déduit un morphisme $\mathbb{Q}_{p} \rightarrow \mathbb{Q}_{p}/\Z_p \rightarrow \mathbb{U}$ de noyau $\mathbb{Z}_{p}$. \\ On note $\psi_{p} : \mathbb{Q}_{p} \rightarrow \mathbb{U}$, il est continu, non trivial et a pour conducteur $\mathbb{Z}_{p}$. \item[Opt 2 :] Le morphisme est donné par le développement de Hensel. \end{enumerate} \end{proof} \begin{lemme} Tout caractère additif $\psi: \mathbb{Q}_{p} \rightarrow \mathbb{C}^*$ est de la forme $x \mapsto \psi_{p}(a x)$ pour un unique $a \in \mathbb{Q}_{p}$. De plus, $\psi$ est non trivial si et seulement si $a \neq 0$, et dans ce cas, le conducteur de $\psi$ est $a^{-1} \mathbb{Z}_{p}$. \end{lemme}
\begin{Théorème}~ \label{thm1} \begin{enumerate} \item[(i)] L'intégrale converge absolument pour tout $y \in F$. \item[(ii)] $\hat{f} \in C_c^{\infty}(F)$. \item[(iii)] $\forall x \in F \quad \widehat{\hat{f}}(x) = f(-x)$ (Formule d'inversion). \end{enumerate} \end{Théorème}
\begin{Théorème}~ \label{thm1} \begin{enumerate} \item[(i)] L'intégrale converge absolument pour tout $y \in F$. \item[(ii)] $\hat{f} \in C_c^{\infty}(F)$. \item[(iii)] $\forall x \in F \quad \widehat{\hat{f}}(x) = f(-x)$ (Formule d'inversion). \end{enumerate} \end{Théorème}
\begin{prop} Soit $\eta : \OO^{\times} \rightarrow \C^*$ un caractère multiplicatif. \begin{enumerate} \item[i)] $\eta$ est localement constant, son noyau contient un sous-groupe ouvert compact $1+\varpi^n \OO$ avec $n$ assez grand. \item[ii)] Im($\eta$) est un sous-groupe fini de $\mathbb{C}^*$ (donc dans $\mathbb{U}_{\infty}$). \end{enumerate} \end{prop}
\begin{prop} Soit $\eta : \OO^{\times} \rightarrow \C^*$ un caractère multiplicatif. \begin{enumerate} \item[i)] $\eta$ est localement constant, son noyau contient un sous-groupe ouvert compact $1+\varpi^n \OO$ avec $n$ assez grand. \item[ii)] Im($\eta$) est un sous-groupe fini de $\mathbb{C}^*$ (donc dans $\mathbb{U}_{\infty}$). \end{enumerate} \end{prop}
\begin{proof}~ \begin{enumerate} \item[(i)] $|f(x) \psi(x y)|_{\infty} = |f(x)|_{\infty}$, donc l'intégrale est absolument convergente (on a même $x \mapsto f(x) \psi(x y)) \in C_c^{\infty}(F)$). \\ L'intégrale qui définit $\hat{f}$ a donc un sens. \item[(ii),(iii)] Par linéarité, il suffit de prouver le théorème pour $f = \1_{a + \varpi^n \OO}$, pour $a \in F$ et $n \in \Z$.\\ Pour $y \in F$, on calcule: $$ \begin{aligned} & \hat{f}(y) = \int_F \1_{a+\varpi^{n} \OO}(x) \psi(x y) \,dx \, {=} \int_F \1_{a+\varpi^{n} \OO}(a+x) \psi((a+x) y) \,dx \\ & = \psi(a y) \int_{\varpi^n \OO} \psi(x y) \,dx. \end{aligned} $$ \end{enumerate} \end{proof}
\begin{proof}~ \begin{enumerate} \item[(i)] $|f(x) \psi(x y)|_{\infty} = |f(x)|_{\infty}$, donc l'intégrale est absolument convergente (on a même $x \mapsto f(x) \psi(x y)) \in C_c^{\infty}(F)$). \\ L'intégrale qui définit $\hat{f}$ a donc un sens. \item[(ii),(iii)] Par linéarité, il suffit de prouver le théorème pour $f = \1_{a + \varpi^n \OO}$, pour $a \in F$ et $n \in \Z$.\\ Pour $y \in F$, on calcule: $$ \begin{aligned} & \hat{f}(y) = \int_F \1_{a+\varpi^{n} \OO}(x) \psi(x y) \,dx \, {=} \int_F \1_{a+\varpi^{n} \OO}(a+x) \psi((a+x) y) \,dx \\ & = \psi(a y) \int_{\varpi^n \OO} \psi(x y) \,dx. \end{aligned} $$ \end{enumerate} \end{proof}
\begin{lemme} L'intégrale converge absolument pour $\Re(s)>0$, et est holomorphe sur ce domaine. \end{lemme} \begin{theorem} Soit $\chi = \eta | \cdot |^s$. Alors, pour toute fonction $f \in \mathcal{S}(F)$, la fonction $\frac{\mathcal{Z}(f, \chi)}{L(\chi)}$ est holomorphe sur $\Re(s) > 0$ et admet un prolongement holomorphe à $\mathbb{C}$ tout entier. \\ De plus, elle satisfait l'équation fonctionnelle suivante : \[ \frac{\mathcal{Z}\left(\hat{f}, \chi^\vee\right)}{L\left(\chi^\vee\right)} = \varepsilon(\chi, \psi) \frac{\mathcal{Z}(f, \chi)}{L(\chi)} \quad \text{ où } \varepsilon(\chi, \psi) = \left\{\begin{array}{ll} i^\varepsilon & \text{si } \eta = (\operatorname{sign})^\varepsilon \; F=\R \\ i^{|n|} & \text{si } \eta = \U \rightarrow \U , \,z \mapsto z^n \; F=\C \end{array}\right. \] \end{theorem}
\begin{lemme} L'intégrale converge absolument pour $\Re(s)>0$, et est holomorphe sur ce domaine. \end{lemme} \begin{theorem} Soit $\chi = \eta | \cdot |^s$. Alors, pour toute fonction $f \in \mathcal{S}(F)$, la fonction $\frac{\mathcal{Z}(f, \chi)}{L(\chi)}$ est holomorphe sur $\Re(s) > 0$ et admet un prolongement holomorphe à $\mathbb{C}$ tout entier. \\ De plus, elle satisfait l'équation fonctionnelle suivante : \[ \frac{\mathcal{Z}\left(\hat{f}, \chi^\vee\right)}{L\left(\chi^\vee\right)} = \varepsilon(\chi, \psi) \frac{\mathcal{Z}(f, \chi)}{L(\chi)} \quad \text{ où } \varepsilon(\chi, \psi) = \left\{\begin{array}{ll} i^\varepsilon & \text{si } \eta = (\operatorname{sign})^\varepsilon \; F=\R \\ i^{|n|} & \text{si } \eta = \U \rightarrow \U , \,z \mapsto z^n \; F=\C \end{array}\right. \] \end{theorem}
\begin{defi} Soit $G$ un groupe abélien localement compact (LC). Un caractère unitaire de $G$ est un morphisme de groupes continu $\chi: G \rightarrow \mathbb{U}$. On note $\hat{G}$ leur groupe, c'est le groupe dual. - On munit $\hat{G}$ de la topologie engendrée par les ensembles $$ W(K, U) = \{\chi \in \hat{G} \mid \chi(K) \subset U\} $$ où $K \subseteq G$ est compact et $U \subseteq \mathbb{C}$ est ouvert.\\ C'est la topologie compacte-ouverte. \end{defi}
\begin{defi} Soit $G$ un groupe abélien localement compact (LC). Un caractère unitaire de $G$ est un morphisme de groupes continu $\chi: G \rightarrow \mathbb{U}$. On note $\hat{G}$ leur groupe, c'est le groupe dual. - On munit $\hat{G}$ de la topologie engendrée par les ensembles $$ W(K, U) = \{\chi \in \hat{G} \mid \chi(K) \subset U\} $$ où $K \subseteq G$ est compact et $U \subseteq \mathbb{C}$ est ouvert.\\ C'est la topologie compacte-ouverte. \end{defi}
Soit $dx$ la mesure de Haar sur $F$ normalisée comme avant. \\ Alors $dx^{\times} = \frac{1}{1 - q^{-1}} \frac{dx}{|x|}$ est une mesure de Haar normalisée sur $F^*$. \begin{lemme} $\operatorname{Vol}(\OO^{\times}, dx^{\times}) = \operatorname{Vol}(\OO, dx) = |\OO/\mathcal{D}|^{-1/2}$ \end{lemme}
Soit $dx$ la mesure de Haar sur $F$ normalisée comme avant. \\ Alors $dx^{\times} = \frac{1}{1 - q^{-1}} \frac{dx}{|x|}$ est une mesure de Haar normalisée sur $F^*$. \begin{lemme} $\operatorname{Vol}(\OO^{\times}, dx^{\times}) = \operatorname{Vol}(\OO, dx) = |\OO/\mathcal{D}|^{-1/2}$ \end{lemme}
\begin{defi} Le conducteur d'un caractère additif $\psi$ non trivial est l'idéal fractionnaire $\varpi^n \OO$ tel que $\varpi^n \OO \subseteq \operatorname{Ker} \psi$ et $\varpi^{n-1} \OO \nsubseteq \operatorname{Ker} \psi$. \\ On le note $\operatorname{Cond}(\psi)$. \end{defi}
\begin{defi} Le conducteur d'un caractère additif $\psi$ non trivial est l'idéal fractionnaire $\varpi^n \OO$ tel que $\varpi^n \OO \subseteq \operatorname{Ker} \psi$ et $\varpi^{n-1} \OO \nsubseteq \operatorname{Ker} \psi$. \\ On le note $\operatorname{Cond}(\psi)$. \end{defi}
\begin{proof}~ Si $f \in C_c^{\infty}(F)$, $f$ ne prend qu'un nombre fini de valeurs. \\ En effet, $\operatorname{supp}(f)=\bigcup\limits_{x \in \text{supp} (f)} U_x$ où $U_x$ est tel que $\left.f\right|_{U_x}$ est constante. Comme $\operatorname{supp}(f)$ est compact, on a un recouvrement fini.\\ Soient $\alpha_1, \ldots, \alpha_n$ les valeurs non nulles prises par $f$. Alors, $f=\sum\limits_{i=1}^n \alpha_i \1_{f^{-1}(\alpha_i)}$, où $f^{-1}(\alpha_i)$ est fermé dans un compact, donc compact.\\ On est donc ramené à prouver la propriété pour $f=\1_K $ où $K$ est compact et ouvert. \\ Pour tout $a \in K$, il existe $n_a$ tel que $(a+\varpi^{n_a} \OO) \subseteq K$. \\ Puisque $K=\bigcup\limits_{a \in K}(a+\varpi^n \OO)$, on peut extraire un sous-recouvrement fini $K=\bigcup\limits_{i=1}^n(a_i+\varpi^{n_i} \OO)$. \\ Ces boules étant disjointes ou incluses l'une dans l'autre, on peut supposer cette union disjointe. Alors, $\1_K=\sum\limits_{i=1}^n \1_{a_i+\varpi^{n_i} \OO}$. \end{proof}
\begin{proof}~ Si $f \in C_c^{\infty}(F)$, $f$ ne prend qu'un nombre fini de valeurs. \\ En effet, $\operatorname{supp}(f)=\bigcup\limits_{x \in \text{supp} (f)} U_x$ où $U_x$ est tel que $\left.f\right|_{U_x}$ est constante. Comme $\operatorname{supp}(f)$ est compact, on a un recouvrement fini.\\ Soient $\alpha_1, \ldots, \alpha_n$ les valeurs non nulles prises par $f$. Alors, $f=\sum\limits_{i=1}^n \alpha_i \1_{f^{-1}(\alpha_i)}$, où $f^{-1}(\alpha_i)$ est fermé dans un compact, donc compact.\\ On est donc ramené à prouver la propriété pour $f=\1_K $ où $K$ est compact et ouvert. \\ Pour tout $a \in K$, il existe $n_a$ tel que $(a+\varpi^{n_a} \OO) \subseteq K$. \\ Puisque $K=\bigcup\limits_{a \in K}(a+\varpi^n \OO)$, on peut extraire un sous-recouvrement fini $K=\bigcup\limits_{i=1}^n(a_i+\varpi^{n_i} \OO)$. \\ Ces boules étant disjointes ou incluses l'une dans l'autre, on peut supposer cette union disjointe. Alors, $\1_K=\sum\limits_{i=1}^n \1_{a_i+\varpi^{n_i} \OO}$. \end{proof}
\begin{proof}~ Pour $\Re(s) > 0$, on a : $$ \begin{aligned} \mathcal{Z}(g, \chi) &= \int_{F^*} g(x) \chi(x) \, dx^{\times} \\ &= \overset{\text{nuls si } k < n_\psi}{\sum_{k=-\infty}^{+\infty}} \int_{\varpi^k \OO^{\times}} \1_{\varpi^{n_\psi} \OO}(x) \psi(x) \overbrace{\chi(x)}^{=|x|^s=q^{-ks}} \, dx^{\times} \\ &= \sum_{k=n_\psi}^{+\infty} q^{-k s} \int_{\varpi^k \OO^{\times}} \, dx^{\times} \\ &= \frac{q^{-n_\psi s}}{1-q^{-s}} \operatorname{Vol}\left(\OO^{\times}, dx^{\times}\right) . \end{aligned} $$ \end{proof} \begin{lemme}~ \begin{aligned}[t] \text{Si } n_\chi=0 \quad \hat{g}&= | \OO / \mathcal{D} |^{-1/2} q^{-n_\psi} \1_{\OO}, \\ \text{Si } n_\chi>0 \quad \hat{g}&= | \OO / \mathcal{D} |^{-1/2} q^{n_\chi-n_\psi} \1_{-U_{n_\chi}}. \end{aligned} \end{lemme} \begin{proof}~ $\hat{g}(y)=\int_F g(x) \psi(x y) d x=\int_{\varpi^{n_\psi-n_\chi} \OO} \psi(x) \psi(x y) d x$ $$ =\int_{\varpi^{n_\psi-n_\chi} \OO} \psi(x(y+1)) d x $$ Or $\operatorname{Cond}(\psi)=\varpi^{n_\psi} \OO$ et $\operatorname{Cond}\left(\psi_{y+1}\right)=(y+1)^{-1} \varpi^{n_\psi} \OO$. \\ L'intégrale est nulle sauf si $\varpi^{n_\psi-n_\chi} \OO \subseteq (y+1)^{-1} \varpi^{n_\psi} \OO$, c'est-à-dire, sauf si $y+1 \in \varpi^{n_\chi} \OO$, c'est-à-dire sauf si $y \in -U_{n_\chi}$. \\ Ainsi \begin{aligned}[t] \hat{g}(y) & =\1_{-U_{n_\chi}}(y) \cdot \operatorname{Vol}\left(\varpi^{n_\psi-n_\chi} \OO, dx\right) \\ & =\1_{-U_{n_\chi}}(y) \cdot|\varpi|^{n_\psi-n_\chi} \quad \operatorname{Vol}(\OO, dx) \\ & =\1_{-U_{n_\chi}}(y) \cdot q^{n_\chi-n_\psi} \cdot| \OO / \mathcal{D} |^{-1/2} \end{aligned} \ \\ \qedhere \end{proof}
\begin{proof}~ Pour $\Re(s) > 0$, on a : $$ \begin{aligned} \mathcal{Z}(g, \chi) &= \int_{F^*} g(x) \chi(x) \, dx^{\times} \\ &= \overset{\text{nuls si } k < n_\psi}{\sum_{k=-\infty}^{+\infty}} \int_{\varpi^k \OO^{\times}} \1_{\varpi^{n_\psi} \OO}(x) \psi(x) \overbrace{\chi(x)}^{=|x|^s=q^{-ks}} \, dx^{\times} \\ &= \sum_{k=n_\psi}^{+\infty} q^{-k s} \int_{\varpi^k \OO^{\times}} \, dx^{\times} \\ &= \frac{q^{-n_\psi s}}{1-q^{-s}} \operatorname{Vol}\left(\OO^{\times}, dx^{\times}\right) . \end{aligned} $$ \end{proof} \begin{lemme}~ \begin{aligned}[t] \text{Si } n_\chi=0 \quad \hat{g}&= | \OO / \mathcal{D} |^{-1/2} q^{-n_\psi} \1_{\OO}, \\ \text{Si } n_\chi>0 \quad \hat{g}&= | \OO / \mathcal{D} |^{-1/2} q^{n_\chi-n_\psi} \1_{-U_{n_\chi}}. \end{aligned} \end{lemme} \begin{proof}~ $\hat{g}(y)=\int_F g(x) \psi(x y) d x=\int_{\varpi^{n_\psi-n_\chi} \OO} \psi(x) \psi(x y) d x$ $$ =\int_{\varpi^{n_\psi-n_\chi} \OO} \psi(x(y+1)) d x $$ Or $\operatorname{Cond}(\psi)=\varpi^{n_\psi} \OO$ et $\operatorname{Cond}\left(\psi_{y+1}\right)=(y+1)^{-1} \varpi^{n_\psi} \OO$. \\ L'intégrale est nulle sauf si $\varpi^{n_\psi-n_\chi} \OO \subseteq (y+1)^{-1} \varpi^{n_\psi} \OO$, c'est-à-dire, sauf si $y+1 \in \varpi^{n_\chi} \OO$, c'est-à-dire sauf si $y \in -U_{n_\chi}$. \\ Ainsi \begin{aligned}[t] \hat{g}(y) & =\1_{-U_{n_\chi}}(y) \cdot \operatorname{Vol}\left(\varpi^{n_\psi-n_\chi} \OO, dx\right) \\ & =\1_{-U_{n_\chi}}(y) \cdot|\varpi|^{n_\psi-n_\chi} \quad \operatorname{Vol}(\OO, dx) \\ & =\1_{-U_{n_\chi}}(y) \cdot q^{n_\chi-n_\psi} \cdot| \OO / \mathcal{D} |^{-1/2} \end{aligned} \ \\ \qedhere \end{proof}
Pour $n \in \mathbb{N}^*$, $\psi\left(\frac{1}{p^n}\right)^{p^n}=\psi(1)=1$, ainsi, $\exists a_n \in \mathbb{Z}$ tel que $\psi\left(\frac{1}{p^n}\right)=e^{\frac{2 i a_n \pi}{p^n}}$. \\ On prend bien sûr $a_0=1$ (car $\psi(1) = 1)$. \\ De plus, $\psi\left(\frac{1}{p^{n+1}}\right)^{p}=\psi\left(\frac{1}{p^n}\right)$, d'où $e^{\frac{2 i a_{n+1} \pi}{p^{n+1}} p}=e^{\frac{2 i a_n \pi}{p^n}}$ et donc $a_{n+1}-a_n \in p^n \mathbb{Z} \subseteq p^n \mathbb{Z}_{p}$. \\ Cela montre que la suite $\left(a_n\right)_n$ est de Cauchy dans $\mathbb{Z}_p$! Elle converge donc vers $a \in \mathbb{Z}_p^*$. On vérifie que $\psi(x)=\psi_p(a x)$ : \begin{itemize} \item C'est clair sur $\mathbb{Z}_p$ car les deux sont triviaux. \item Il reste à vérifier sur les $\frac{1}{p^n}, n \in \mathbb{N}^*$, mais : $$ \psi\left(\frac{1}{p^n}\right)=e^{\frac{2 i a_n \pi}{p^n}} \text{ et } \frac{a}{p^n} \in \frac{a_n}{p^n}+\mathbb{Z}_p, \text{ donc } \psi_{p}\left(\frac{a}{p^n}\right)=\psi\left(\frac{1}{p^n}\right). $$ \end{itemize} \end{proof}
Pour $n \in \mathbb{N}^*$, $\psi\left(\frac{1}{p^n}\right)^{p^n}=\psi(1)=1$, ainsi, $\exists a_n \in \mathbb{Z}$ tel que $\psi\left(\frac{1}{p^n}\right)=e^{\frac{2 i a_n \pi}{p^n}}$. \\ On prend bien sûr $a_0=1$ (car $\psi(1) = 1)$. \\ De plus, $\psi\left(\frac{1}{p^{n+1}}\right)^{p}=\psi\left(\frac{1}{p^n}\right)$, d'où $e^{\frac{2 i a_{n+1} \pi}{p^{n+1}} p}=e^{\frac{2 i a_n \pi}{p^n}}$ et donc $a_{n+1}-a_n \in p^n \mathbb{Z} \subseteq p^n \mathbb{Z}_{p}$. \\ Cela montre que la suite $\left(a_n\right)_n$ est de Cauchy dans $\mathbb{Z}_p$! Elle converge donc vers $a \in \mathbb{Z}_p^*$. On vérifie que $\psi(x)=\psi_p(a x)$ : \begin{itemize} \item C'est clair sur $\mathbb{Z}_p$ car les deux sont triviaux. \item Il reste à vérifier sur les $\frac{1}{p^n}, n \in \mathbb{N}^*$, mais : $$ \psi\left(\frac{1}{p^n}\right)=e^{\frac{2 i a_n \pi}{p^n}} \text{ et } \frac{a}{p^n} \in \frac{a_n}{p^n}+\mathbb{Z}_p, \text{ donc } \psi_{p}\left(\frac{a}{p^n}\right)=\psi\left(\frac{1}{p^n}\right). $$ \end{itemize} \end{proof}
\underline{Continuité de $\theta$} : \\ Il suffit de voir que $\theta^{-1}(W(K, N(1)))$ est un voisinage de $0$ dans $F$, pour les $K \subseteq F$ compact. \\ Or, $\theta^{-1}(W(K, N(1)))=\left\{a \in F \mid \psi_0(a K) \subseteq N(1)\right\}. $ Comme $K$ est compact, on peut trouver $\varepsilon > 0$ tel que $B(0, \varepsilon) \cdot K \subseteq V$, où $V$ est un voisinage de $0$ dans $F$ tel que $\psi_0(V) \subseteq N(1)$ (existe par continuité de $\psi_0$). \\ On a alors $B(0, \varepsilon) \subseteq \theta^{-1}(W(K, N(1)))$. \\
\underline{Continuité de $\theta$} : \\ Il suffit de voir que $\theta^{-1}(W(K, N(1)))$ est un voisinage de $0$ dans $F$, pour les $K \subseteq F$ compact. \\ Or, $\theta^{-1}(W(K, N(1)))=\left\{a \in F \mid \psi_0(a K) \subseteq N(1)\right\}. $ Comme $K$ est compact, on peut trouver $\varepsilon > 0$ tel que $B(0, \varepsilon) \cdot K \subseteq V$, où $V$ est un voisinage de $0$ dans $F$ tel que $\psi_0(V) \subseteq N(1)$ (existe par continuité de $\psi_0$). \\ On a alors $B(0, \varepsilon) \subseteq \theta^{-1}(W(K, N(1)))$. \\
\begin{rmk}~ \begin{enumerate} \item[1)] Cette décomposition dépend fortement du choix de $\varpi$. \item[2)] $\chi |_{\OO^{\times}}=\eta |_{\OO^{\times}}$, $\eta$ est donc déterminé par $\chi$. \item[3)] $\chi(\varpi)=\mid \varpi\mid^s=q^{-s}=e^{-s \log q}$. \\ $s$ est donc bien défini modulo $\frac{2 i \pi}{\log q} \mathbb{Z}$. En particulier, $\Re(s)$ est bien définie. \end{enumerate} \end{rmk}
\begin{rmk}~ \begin{enumerate} \item[1)] Cette décomposition dépend fortement du choix de $\varpi$. \item[2)] $\chi |_{\OO^{\times}}=\eta |_{\OO^{\times}}$, $\eta$ est donc déterminé par $\chi$. \item[3)] $\chi(\varpi)=\mid \varpi\mid^s=q^{-s}=e^{-s \log q}$. \\ $s$ est donc bien défini modulo $\frac{2 i \pi}{\log q} \mathbb{Z}$. En particulier, $\Re(s)$ est bien définie. \end{enumerate} \end{rmk}
Pour l'inclusion réciproque, on va montrer que $\forall \chi \in W , \; \forall K_1 \subseteq G$ compact $\forall m \geq 1 , \\ W(\chi) := W \cap \chi \cdot W(K_1, N(1 / m))$ est un voisinage de $\chi$ pour $\tau_0$.\\ \ \\ $K$ étant un voisinage compact de $e$ dans $G$, $\exists V$ voisinage ouvert de $e$ dans $G$ tel que $V^{(2 m)} \subseteq K$. \\ Par compacité de $K_1$, $\exists F$ ensemble fini de $K_1$ tel que $K_1 \subseteq F V$. Posons $W_0(\chi) := W_0 \cap \chi \cdot W_0(F, N(1 / 2 m))$. C'est un ouvert de $\tau_0$ qui contient $\chi$. Il suffit de voir que $W_0(\chi) \subseteq W(\chi)$. Soit $\mu \in W_0(\chi) \subseteq W_0 = W. $ On a $ \mu = \chi \mu_0$ avec $\mu_0 \in W_0(F, N(1 / 2 m))$. \\ $\mu_0 \in \hat{G}_0$ mais $\mu_0 = \mu \chi^{-1} \in \widehat{G} \cdot \widehat{G} = \widehat{G}$ donc $\mu_0$ est continue pour la topologie de $G$. $$ \mu_0(K) \subseteq \mu(K) \chi^{-1}(K) \subseteq \overline{N(1 / 4)} \overline{N(1 / 4)} \subseteq N(1) \quad \text{d'où } \mu_0 \in W(K, N(1)). $$ \\ Par conséquent, $\mu_0\left(V^{(2n)}\right) \subseteq N(1)$ et par le lemme, $\mu_0(V) \subseteq N(1 / 2 m)$. \\ Donc, $\mu_0\left(K_1\right) \subseteq \mu_0(F \cdot V) \subseteq N(1 / 2 m) \cdot \mu_0(V) \subseteq N(1 / 2 m) N(1 / 2 m) \subseteq N(1 / m)$. Ainsi, $\mu_0 \in W\left(K_1, N(1 / m)\right)$ et $\mu = \chi \mu_0 \in W(\chi)$. \end{enumerate} \end{proof}
Pour l'inclusion réciproque, on va montrer que $\forall \chi \in W , \; \forall K_1 \subseteq G$ compact $\forall m \geq 1 , \\ W(\chi) := W \cap \chi \cdot W(K_1, N(1 / m))$ est un voisinage de $\chi$ pour $\tau_0$.\\ \ \\ $K$ étant un voisinage compact de $e$ dans $G$, $\exists V$ voisinage ouvert de $e$ dans $G$ tel que $V^{(2 m)} \subseteq K$. \\ Par compacité de $K_1$, $\exists F$ ensemble fini de $K_1$ tel que $K_1 \subseteq F V$. Posons $W_0(\chi) := W_0 \cap \chi \cdot W_0(F, N(1 / 2 m))$. C'est un ouvert de $\tau_0$ qui contient $\chi$. Il suffit de voir que $W_0(\chi) \subseteq W(\chi)$. Soit $\mu \in W_0(\chi) \subseteq W_0 = W. $ On a $ \mu = \chi \mu_0$ avec $\mu_0 \in W_0(F, N(1 / 2 m))$. \\ $\mu_0 \in \hat{G}_0$ mais $\mu_0 = \mu \chi^{-1} \in \widehat{G} \cdot \widehat{G} = \widehat{G}$ donc $\mu_0$ est continue pour la topologie de $G$. $$ \mu_0(K) \subseteq \mu(K) \chi^{-1}(K) \subseteq \overline{N(1 / 4)} \overline{N(1 / 4)} \subseteq N(1) \quad \text{d'où } \mu_0 \in W(K, N(1)). $$ \\ Par conséquent, $\mu_0\left(V^{(2n)}\right) \subseteq N(1)$ et par le lemme, $\mu_0(V) \subseteq N(1 / 2 m)$. \\ Donc, $\mu_0\left(K_1\right) \subseteq \mu_0(F \cdot V) \subseteq N(1 / 2 m) \cdot \mu_0(V) \subseteq N(1 / 2 m) N(1 / 2 m) \subseteq N(1 / m)$. Ainsi, $\mu_0 \in W\left(K_1, N(1 / m)\right)$ et $\mu = \chi \mu_0 \in W(\chi)$. \end{enumerate} \end{proof}
\begin{prop} \label{prop1} Tout caractère multiplicatif $\chi:F^{*} \rightarrow \mathbb{C}^*$ s'écrit $\chi=\eta \cdot |.|^s$ où $\eta : \OO^{\times} \rightarrow \mathbb{C}^*$ est un caractère étendu à $F^{*}$ par $\eta(\varpi)=1$ et $s \in \mathbb{C}$. \end{prop}
\begin{prop} \label{prop1} Tout caractère multiplicatif $\chi:F^{*} \rightarrow \mathbb{C}^*$ s'écrit $\chi=\eta \cdot |.|^s$ où $\eta : \OO^{\times} \rightarrow \mathbb{C}^*$ est un caractère étendu à $F^{*}$ par $\eta(\varpi)=1$ et $s \in \mathbb{C}$. \end{prop}
\begin{lemme} On a un isomorphisme de groupes additifs: $$ \mathbb{Z}[1 / p] / \mathbb{Z} \simeq \mathbb{Q}_p / \mathbb{Z}_p $$ \end{lemme}
\begin{lemme} On a un isomorphisme de groupes additifs: $$ \mathbb{Z}[1 / p] / \mathbb{Z} \simeq \mathbb{Q}_p / \mathbb{Z}_p $$ \end{lemme}
\begin{EX} \begin{enumerate} \item[1)] La $v.a$ normalisée sur $f$ n'est pas à support compact. \item[2)] $x \mapsto |x| \cdot \1_{\mathbb{Z}_p}(x)$ n'est pas localement constante en $0$. \item[3)] Si $f \in C_c^{\infty}(F)$, alors $|f|_{\infty} \in C_c^{\infty}(F)$. $f$ est donc continue, à support compact, et on a $f \in L^1(F, dx)$ . \end{enumerate} \end{EX}
\begin{EX} \begin{enumerate} \item[1)] La $v.a$ normalisée sur $f$ n'est pas à support compact. \item[2)] $x \mapsto |x| \cdot \1_{\mathbb{Z}_p}(x)$ n'est pas localement constante en $0$. \item[3)] Si $f \in C_c^{\infty}(F)$, alors $|f|_{\infty} \in C_c^{\infty}(F)$. $f$ est donc continue, à support compact, et on a $f \in L^1(F, dx)$ . \end{enumerate} \end{EX}
\begin{Cor} Le seul sous-groupe de $\mathbb{C}^*$ inclus dans $N(1)$ est trivial. \end{Cor}
\begin{Cor} Le seul sous-groupe de $\mathbb{C}^*$ inclus dans $N(1)$ est trivial. \end{Cor}
\begin{EX} \begin{enumerate} \item Les caractères multiplicatifs sont localement constants, mais pas à support compact (par exemple, $x \mapsto |x|$). \item Pour $f \in C_c^{\infty}(F)$, $f|_{F^*}$ est localement constante.\\ ~- Si $f(0) = 0$, alors $\exists U$ voisinage de $0$ avec $f(U) = 0$. \\ Dans ce cas, $\operatorname{supp}(f) \subseteq F \setminus U \subseteq F^*$, d'où $\operatorname{supp}(f|_{F^*}) = \operatorname{supp}(f)$ est compact.\\ ~- Réciproquement, si $\operatorname{supp}(f|_{F^*})$ est compact, $x \mapsto |x|$ est bornée et atteint ses bornes sur le compact $\operatorname{supp}(f) \cap F^* = \operatorname{supp}(f|_{F^*})$ : $$\exists m, M > 0 \quad \forall x \in \operatorname{supp}(f|_{F^*}) \quad m \leq |x| \leq M.$$ Ainsi, $f|_{B(0, m) \setminus \{0\}} = 0$, et $f(0) = 0$ donc $f$ est localement constante. \\ Ainsi, pour $f \in C_c^{\infty}(F)$, on a : $$f_{C^*} \in C_{F^*}^{\infty}(F^*) \Leftrightarrow f(0) = 0.$$ Par exemple, $\1_{\mathbb{Z}_{p}} \in C_c^{\infty}(\mathbb{Q}_{p})$ mais n'appartient pas à $C_c^{\infty}(\Q_{p}^*)$. \end{enumerate} \end{EX}
\begin{EX} \begin{enumerate} \item Les caractères multiplicatifs sont localement constants, mais pas à support compact (par exemple, $x \mapsto |x|$). \item Pour $f \in C_c^{\infty}(F)$, $f|_{F^*}$ est localement constante.\\ ~- Si $f(0) = 0$, alors $\exists U$ voisinage de $0$ avec $f(U) = 0$. \\ Dans ce cas, $\operatorname{supp}(f) \subseteq F \setminus U \subseteq F^*$, d'où $\operatorname{supp}(f|_{F^*}) = \operatorname{supp}(f)$ est compact.\\ ~- Réciproquement, si $\operatorname{supp}(f|_{F^*})$ est compact, $x \mapsto |x|$ est bornée et atteint ses bornes sur le compact $\operatorname{supp}(f) \cap F^* = \operatorname{supp}(f|_{F^*})$ : $$\exists m, M > 0 \quad \forall x \in \operatorname{supp}(f|_{F^*}) \quad m \leq |x| \leq M.$$ Ainsi, $f|_{B(0, m) \setminus \{0\}} = 0$, et $f(0) = 0$ donc $f$ est localement constante. \\ Ainsi, pour $f \in C_c^{\infty}(F)$, on a : $$f_{C^*} \in C_{F^*}^{\infty}(F^*) \Leftrightarrow f(0) = 0.$$ Par exemple, $\1_{\mathbb{Z}_{p}} \in C_c^{\infty}(\mathbb{Q}_{p})$ mais n'appartient pas à $C_c^{\infty}(\Q_{p}^*)$. \end{enumerate} \end{EX}
\begin{proof}~ \begin{enumerate} \item[(i)] Si $G$ est discret, $\hat{G} \subseteq \mathbb{U}^G$ est muni de la topologie induite par la topologie produit sur $\mathbb{U}^G$. Or, ce dernier est compact. Il suffit donc de montrer que $\hat{G} \subseteq \mathbb{U}^G$ est fermé. Soit $\phi \in \mathbb{U}^G$ qui est dans l'adhérence de $\hat{G}$. Soit $\varepsilon > 0$, pour $\forall x \in G$, définissons \\ $B_x := \left\{\psi \in \mathbb{U}^G \mid |\psi(x)-\phi(x)| < \varepsilon\right\}$ est un ouvert de $\U^G$ qui contient $\phi$. Pour $(x, y) \in G^2$, soit $\chi \in B_x \cap B_y \cap B_{xy} \cap \hat{G}$ qui est un ouvert non vide ($\phi$ est dans l'adhérence de $\hat{G}$). On a : \begin{align*} |\phi(x)\phi(y)-\phi(xy)| &\leqslant |\phi(x)-\chi(x)|\cdot|\phi(y)| + |\chi(x)|\cdot|\phi(y)-\chi(y)| + |\chi(xy)-\phi(xy)| \\ &< 3\varepsilon. \end{align*} Puisque $\varepsilon$ est arbitraire, on conclut que $\phi(xy) = \phi(x)\phi(y)$. Ainsi, $\phi \in \hat{G}$ (et $\phi$ est bien continue car $G$ est discret) et donc $\hat{G} \subseteq \mathbb{U}^G$ est fermé. \item[(ii)] Si $G$ est compact, $W(G, N(1))$ est un voisinage compact de $\1$. Pour tout $\chi \in W(G, N(1))$, $\chi(G)$ est un sous-groupe de $\mathbb{C}^*$ inclus dans $N(1)$, donc trivial. Ainsi, $\{\1\} = W(G, N(1))$ est ouvert, et $\hat{G}$ est discret. \item[(iii)] Il suffit de voir que les $W(K, \overline{N(1 / 4)})$, où $K$ est un voisinage compact de $e$, sont compacts dans $\hat{G}$ (ils forment une base de voisinage de $\1$). On note $G_0$ le groupe topologique discret ayant le même groupe sous-jacent que $G$ et on fixe un voisinage compact $K$ de $e$ dans $G$. On considère $W_0 = \{\chi \in \hat{G}_0 \mid \chi(K) \subseteq \overline{N(1 / 4)}\}$ et $W = W(K, \overline{N(1 / 4)})$ et on veut montrer que $W$ est compact. $W_0$ l'est, car il est fermé dans $\hat{G}_0$ qui est compact par (i).\\ \ \\ Bien sûr, $W \subseteq W_0$. Mais on a aussi $W_0 \subseteq W$, car si $K \subseteq \chi^{-1}(\overline{N(1 / 4)}) \subseteq \chi^{-1}(N(1))$, cela montre que $\chi^{-1}(N(1))$ est un voisinage de $e \in G$, le lemme dit que $\chi$ est continue, donc $\chi \in \widehat{G}$. \\ \ \\ Ainsi ensemblistement $W=W_0$. On a deux topologies sur $W$, issues de $\hat{G}$ et $\hat{G}_0$. Montrons qu'elles sont identiques: \begin{itemize} \item La topologie sur $\hat{G}$ est engendrée par les ensembles : $$ \mathcal{W}(K, U) := \{\chi \in \hat{G} \mid \chi(K) \subseteq U\}, \quad K \subseteq G \text{ compact, } U \subseteq \mathbb{U} \text{ ouvert} $$ \item La topologie sur $\hat{G}_0$ est engendrée par les ensembles : $$ W_0(K_0, U_0) = \{\chi \in \hat{G}_0 \mid \chi(K_0) \subseteq U_0\}, \quad K_0 \subseteq G_0 \text{ compact (ie fini) }, U_0 \subseteq \U \text{ ouvert} $$ \end{itemize}
\begin{proof}~ \begin{enumerate} \item[(i)] Si $G$ est discret, $\hat{G} \subseteq \mathbb{U}^G$ est muni de la topologie induite par la topologie produit sur $\mathbb{U}^G$. Or, ce dernier est compact. Il suffit donc de montrer que $\hat{G} \subseteq \mathbb{U}^G$ est fermé. Soit $\phi \in \mathbb{U}^G$ qui est dans l'adhérence de $\hat{G}$. Soit $\varepsilon > 0$, pour $\forall x \in G$, définissons \\ $B_x := \left\{\psi \in \mathbb{U}^G \mid |\psi(x)-\phi(x)| < \varepsilon\right\}$ est un ouvert de $\U^G$ qui contient $\phi$. Pour $(x, y) \in G^2$, soit $\chi \in B_x \cap B_y \cap B_{xy} \cap \hat{G}$ qui est un ouvert non vide ($\phi$ est dans l'adhérence de $\hat{G}$). On a : \begin{align*} |\phi(x)\phi(y)-\phi(xy)| &\leqslant |\phi(x)-\chi(x)|\cdot|\phi(y)| + |\chi(x)|\cdot|\phi(y)-\chi(y)| + |\chi(xy)-\phi(xy)| \\ &< 3\varepsilon. \end{align*} Puisque $\varepsilon$ est arbitraire, on conclut que $\phi(xy) = \phi(x)\phi(y)$. Ainsi, $\phi \in \hat{G}$ (et $\phi$ est bien continue car $G$ est discret) et donc $\hat{G} \subseteq \mathbb{U}^G$ est fermé. \item[(ii)] Si $G$ est compact, $W(G, N(1))$ est un voisinage compact de $\1$. Pour tout $\chi \in W(G, N(1))$, $\chi(G)$ est un sous-groupe de $\mathbb{C}^*$ inclus dans $N(1)$, donc trivial. Ainsi, $\{\1\} = W(G, N(1))$ est ouvert, et $\hat{G}$ est discret. \item[(iii)] Il suffit de voir que les $W(K, \overline{N(1 / 4)})$, où $K$ est un voisinage compact de $e$, sont compacts dans $\hat{G}$ (ils forment une base de voisinage de $\1$). On note $G_0$ le groupe topologique discret ayant le même groupe sous-jacent que $G$ et on fixe un voisinage compact $K$ de $e$ dans $G$. On considère $W_0 = \{\chi \in \hat{G}_0 \mid \chi(K) \subseteq \overline{N(1 / 4)}\}$ et $W = W(K, \overline{N(1 / 4)})$ et on veut montrer que $W$ est compact. $W_0$ l'est, car il est fermé dans $\hat{G}_0$ qui est compact par (i).\\ \ \\ Bien sûr, $W \subseteq W_0$. Mais on a aussi $W_0 \subseteq W$, car si $K \subseteq \chi^{-1}(\overline{N(1 / 4)}) \subseteq \chi^{-1}(N(1))$, cela montre que $\chi^{-1}(N(1))$ est un voisinage de $e \in G$, le lemme dit que $\chi$ est continue, donc $\chi \in \widehat{G}$. \\ \ \\ Ainsi ensemblistement $W=W_0$. On a deux topologies sur $W$, issues de $\hat{G}$ et $\hat{G}_0$. Montrons qu'elles sont identiques: \begin{itemize} \item La topologie sur $\hat{G}$ est engendrée par les ensembles : $$ \mathcal{W}(K, U) := \{\chi \in \hat{G} \mid \chi(K) \subseteq U\}, \quad K \subseteq G \text{ compact, } U \subseteq \mathbb{U} \text{ ouvert} $$ \item La topologie sur $\hat{G}_0$ est engendrée par les ensembles : $$ W_0(K_0, U_0) = \{\chi \in \hat{G}_0 \mid \chi(K_0) \subseteq U_0\}, \quad K_0 \subseteq G_0 \text{ compact (ie fini) }, U_0 \subseteq \U \text{ ouvert} $$ \end{itemize}
\begin{lemme} La topologie compacte-ouverte est engendrée par les $W(K, B(\alpha, \varepsilon))$ avec $K \subseteq G$ compact, $\alpha \in \U $, $\varepsilon > 0$. \end{lemme}
\begin{lemme} La topologie compacte-ouverte est engendrée par les $W(K, B(\alpha, \varepsilon))$ avec $K \subseteq G$ compact, $\alpha \in \U $, $\varepsilon > 0$. \end{lemme}
\begin{lemme} Soit $G$ un groupe topologique localement compact. \begin{enumerate} \item[(i)] Un morphisme de groupe $\chi: G \rightarrow \U$ est continu si et seulement si $\chi^{-1}(N(1))$ est un voisinage de $e:=1_G$. \item[(ii)] La famille des $W(K, N(1))$ indexée par les compacts $K$ de $G$ est une base de voisinages de $1$. \end{enumerate} \end{lemme}
\begin{lemme} Soit $G$ un groupe topologique localement compact. \begin{enumerate} \item[(i)] Un morphisme de groupe $\chi: G \rightarrow \U$ est continu si et seulement si $\chi^{-1}(N(1))$ est un voisinage de $e:=1_G$. \item[(ii)] La famille des $W(K, N(1))$ indexée par les compacts $K$ de $G$ est une base de voisinages de $1$. \end{enumerate} \end{lemme}
\begin{proof}~ \underline{Unicité} : Si $a, b \in \mathbb{Q}_{p}$ vérifient $\forall x \in \mathbb{Q}_{p} \quad \psi_{p}(a x)=\psi_{p}(b x)$, alors pour tout $x \in \mathbb{Q}_{p}$, $\psi_{p}((a-b)x)=0$, c'est-à-dire $(a-b) x \in \mathbb{Z}_{p}$. Ceci implique $a-b=0$, sans quoi $(a-b) \cdot \varpi^{-v_{p}(a-b)-1} \notin \mathbb{Z}_{p}$.\\ \underline{Existence} : Soit $\psi$ un caractère additif. $\mathbb{Q}_{p}=\bigcup\limits_{n \geq 0} p^{-n} \mathbb{Z}_{p}$, et $\left.\psi\right|_{p^{-n} \mathbb{Z}_{p}}$ est déterminé par $\psi(1 / p^n)$. \\ Si $\psi$ est trivial, $a=0$ convient. \\ Sinon, en remplaçant $\psi$ par $\psi_{b}: x \mapsto \psi(b x)$, on a $\operatorname{Cond}(\psi_{b})=b^{-1}\operatorname{Cond}(\psi)$. On peut donc se contenter de traiter le cas $\operatorname{Cond}(\psi)=\mathbb{Z}_{p}$.
\begin{proof}~ \underline{Unicité} : Si $a, b \in \mathbb{Q}_{p}$ vérifient $\forall x \in \mathbb{Q}_{p} \quad \psi_{p}(a x)=\psi_{p}(b x)$, alors pour tout $x \in \mathbb{Q}_{p}$, $\psi_{p}((a-b)x)=0$, c'est-à-dire $(a-b) x \in \mathbb{Z}_{p}$. Ceci implique $a-b=0$, sans quoi $(a-b) \cdot \varpi^{-v_{p}(a-b)-1} \notin \mathbb{Z}_{p}$.\\ \underline{Existence} : Soit $\psi$ un caractère additif. $\mathbb{Q}_{p}=\bigcup\limits_{n \geq 0} p^{-n} \mathbb{Z}_{p}$, et $\left.\psi\right|_{p^{-n} \mathbb{Z}_{p}}$ est déterminé par $\psi(1 / p^n)$. \\ Si $\psi$ est trivial, $a=0$ convient. \\ Sinon, en remplaçant $\psi$ par $\psi_{b}: x \mapsto \psi(b x)$, on a $\operatorname{Cond}(\psi_{b})=b^{-1}\operatorname{Cond}(\psi)$. On peut donc se contenter de traiter le cas $\operatorname{Cond}(\psi)=\mathbb{Z}_{p}$.
\begin{rmk} Dans un cadre global, pour un corps de nombres $K / \mathbb{Q}$, on a donc, pour presque tout $p$ : $\forall v \mid p$, la forme $\alpha=\operatorname{Tr}_{K_v / \Q_p} p$ est à déterminant (dans une $\mathbb{Z}_{p}$-base de $\OO_{K_v}$) dans $\OO_{K_v}^*$. Dans ce cas, $\mathcal{D}_{K_v / \mathbb{Q}_p}=\OO_{K_v}^*$. \end{rmk}
\begin{rmk} Dans un cadre global, pour un corps de nombres $K / \mathbb{Q}$, on a donc, pour presque tout $p$ : $\forall v \mid p$, la forme $\alpha=\operatorname{Tr}_{K_v / \Q_p} p$ est à déterminant (dans une $\mathbb{Z}_{p}$-base de $\OO_{K_v}$) dans $\OO_{K_v}^*$. Dans ce cas, $\mathcal{D}_{K_v / \mathbb{Q}_p}=\OO_{K_v}^*$. \end{rmk}
\begin{proof}~ Par le lemme précédent, un voisinage $V$ de $\chi$ contient un $\bigcap\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon_i\right)\right) \ni \chi$. \\ On note $m_i := \sup\limits_{k \in K_i} |\chi(k_i) - \alpha_i| < \varepsilon_i$. Prenons $0 < \eta < \min\limits_{1 \leq i \leq n} \frac{\varepsilon_i - m_i}{2}$. \\ Alors $B_K(\chi, \eta) \subseteq \bigcap\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon_i\right)\right) \subseteq V$ car pour $ \phi \in B_K(\chi, \eta)$, $\forall i , \forall k \in K_i \\ \quad |\phi(k) - \alpha_i| \leq |\phi(k) - \chi(k)| + |\chi(k) - \alpha_i| \leq \eta + m_i \leq \frac{\varepsilon_i + m_i}{2} < \varepsilon_i$. \\ Prouvons maintenant que $B_K(\chi, \varepsilon)$ est un voisinage de $\chi$. Par compacité de $K$, $\exists \alpha_1, \ldots, \alpha_n$ tels que $\chi(K) \subseteq \bigcup\limits_{i=1}^n B\left(\alpha_i, \varepsilon / 2\right)$. Alors, $K = \bigcup\limits_{i=1}^n K \cap \chi^{-1}\left(B\left(\alpha_i, \varepsilon / 2\right)\right)$, donc par normalité de $K$, pour tout $i$, il existe $K_i$ compact avec $K_i \subseteq K \cap \chi^{-1}\left(B\left(\alpha_i, \varepsilon / 2\right)\right)$ tels que $K = \bigcup\limits_{i=1}^n K_i$. Ainsi, $\chi \in \bigcap\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon / 2\right)\right)$. \\ Soit $\phi$ dans cet ouvert. Pour tout $k \in K_i$, $|\phi(k) - \chi(k)| \leqslant |\phi(k) - \alpha_i| + |\chi(k) - \alpha_i| < \frac{\varepsilon}{2} + \frac{\varepsilon}{2} = \varepsilon$. D'où $\bigcap\limits_{i=1}^{n} W\left(K_i, B\left(\alpha_i, \varepsilon / 2\right)\right) \subseteq B_K(\chi, \varepsilon)$. \end{proof}
\begin{proof}~ Par le lemme précédent, un voisinage $V$ de $\chi$ contient un $\bigcap\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon_i\right)\right) \ni \chi$. \\ On note $m_i := \sup\limits_{k \in K_i} |\chi(k_i) - \alpha_i| < \varepsilon_i$. Prenons $0 < \eta < \min\limits_{1 \leq i \leq n} \frac{\varepsilon_i - m_i}{2}$. \\ Alors $B_K(\chi, \eta) \subseteq \bigcap\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon_i\right)\right) \subseteq V$ car pour $ \phi \in B_K(\chi, \eta)$, $\forall i , \forall k \in K_i \\ \quad |\phi(k) - \alpha_i| \leq |\phi(k) - \chi(k)| + |\chi(k) - \alpha_i| \leq \eta + m_i \leq \frac{\varepsilon_i + m_i}{2} < \varepsilon_i$. \\ Prouvons maintenant que $B_K(\chi, \varepsilon)$ est un voisinage de $\chi$. Par compacité de $K$, $\exists \alpha_1, \ldots, \alpha_n$ tels que $\chi(K) \subseteq \bigcup\limits_{i=1}^n B\left(\alpha_i, \varepsilon / 2\right)$. Alors, $K = \bigcup\limits_{i=1}^n K \cap \chi^{-1}\left(B\left(\alpha_i, \varepsilon / 2\right)\right)$, donc par normalité de $K$, pour tout $i$, il existe $K_i$ compact avec $K_i \subseteq K \cap \chi^{-1}\left(B\left(\alpha_i, \varepsilon / 2\right)\right)$ tels que $K = \bigcup\limits_{i=1}^n K_i$. Ainsi, $\chi \in \bigcap\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon / 2\right)\right)$. \\ Soit $\phi$ dans cet ouvert. Pour tout $k \in K_i$, $|\phi(k) - \chi(k)| \leqslant |\phi(k) - \alpha_i| + |\chi(k) - \alpha_i| < \frac{\varepsilon}{2} + \frac{\varepsilon}{2} = \varepsilon$. D'où $\bigcap\limits_{i=1}^{n} W\left(K_i, B\left(\alpha_i, \varepsilon / 2\right)\right) \subseteq B_K(\chi, \varepsilon)$. \end{proof}
Notons $\tau$ (resp. $\tau_0$) la topologie sur $W = W_0$ induite par celle de $\hat{G}$ (resp. $\hat{G}_0$). On souhaite montrer $\tau = \tau_0$.\\ Les compacts de $G_0$ sont finis, donc compacts pour $G$, d'où $\tau_0 \subset \tau$. \\
Notons $\tau$ (resp. $\tau_0$) la topologie sur $W = W_0$ induite par celle de $\hat{G}$ (resp. $\hat{G}_0$). On souhaite montrer $\tau = \tau_0$.\\ Les compacts de $G_0$ sont finis, donc compacts pour $G$, d'où $\tau_0 \subset \tau$. \\
\begin{lemme} \label{lem} Soit $m \geqslant 1$. Soit $1_G \in V \subseteq G$. Pour tout homomorphisme $\chi: G \rightarrow \mathbb{U}$ tel que $\chi\left(V^{(m)}\right) \subseteq N(1)$, on a $\chi(V) \subseteq N\left(\frac{1}{m}\right)$. \end{lemme}
\begin{lemme} \label{lem} Soit $m \geqslant 1$. Soit $1_G \in V \subseteq G$. Pour tout homomorphisme $\chi: G \rightarrow \mathbb{U}$ tel que $\chi\left(V^{(m)}\right) \subseteq N(1)$, on a $\chi(V) \subseteq N\left(\frac{1}{m}\right)$. \end{lemme}
\begin{rmk}~ \begin{enumerate} \item[1)] $|\chi(\varpi)|=1 \Leftrightarrow \operatorname{Re}(s)=0$, (car $\chi(\varpi)=|\varpi|^s=e^{-s \log q}$). \item[2)] $\OO^{\times}$ est compact donc $\widehat{\OO^{\times}}$ est discret.\\ De $F^*=\OO^{\times} \times \mathbb{Z}$, on tire $\widehat{F^*}=\widehat{\OO^{\times}} \times \underbrace{i \mathbb{R} / 2 i \pi \log q}_{\text{Compact, à voir comme $\hat{\mathbb{Z}}=\mathbb{U}$}}$. \end{enumerate} Ainsi, l'ensemble des caractères de $F^*$ (avec aussi ceux qui ne sont pas unitaires (ramifiés ?) ) est une réunion disjointe indexée par $\OO^{\times}$ de $\mathbb{C} / 2 i \pi \log q$. \\ On a une action de $\mathbb{C}$ sur les caractères donnée par: $$ s \in \mathbb{C}: \chi \mapsto \chi | \cdot |^s $$ \end{rmk}
\begin{rmk}~ \begin{enumerate} \item[1)] $|\chi(\varpi)|=1 \Leftrightarrow \operatorname{Re}(s)=0$, (car $\chi(\varpi)=|\varpi|^s=e^{-s \log q}$). \item[2)] $\OO^{\times}$ est compact donc $\widehat{\OO^{\times}}$ est discret.\\ De $F^*=\OO^{\times} \times \mathbb{Z}$, on tire $\widehat{F^*}=\widehat{\OO^{\times}} \times \underbrace{i \mathbb{R} / 2 i \pi \log q}_{\text{Compact, à voir comme $\hat{\mathbb{Z}}=\mathbb{U}$}}$. \end{enumerate} Ainsi, l'ensemble des caractères de $F^*$ (avec aussi ceux qui ne sont pas unitaires (ramifiés ?) ) est une réunion disjointe indexée par $\OO^{\times}$ de $\mathbb{C} / 2 i \pi \log q$. \\ On a une action de $\mathbb{C}$ sur les caractères donnée par: $$ s \in \mathbb{C}: \chi \mapsto \chi | \cdot |^s $$ \end{rmk}
\begin{defi} $C_c^{\infty}(F^*)$ est l'espace des applications $F^* \longrightarrow \mathbb{C}$ localement constantes sur $F^*$ à support compact. \end{defi}
\begin{defi} $C_c^{\infty}(F^*)$ est l'espace des applications $F^* \longrightarrow \mathbb{C}$ localement constantes sur $F^*$ à support compact. \end{defi}
\begin{rmk}~ \begin{enumerate} \item[1)] La condition $(iii)$ détermine la mesure $dx$ sur $F$. Une mesure qui vérifie $(iii)$ est dite autoduale pour le caractère $\psi$. \item[2)] Rappel: $\operatorname{cond}(\psi) = \mathcal{D}^{-1}$. \end{enumerate} \end{rmk}
\begin{rmk}~ \begin{enumerate} \item[1)] La condition $(iii)$ détermine la mesure $dx$ sur $F$. Une mesure qui vérifie $(iii)$ est dite autoduale pour le caractère $\psi$. \item[2)] Rappel: $\operatorname{cond}(\psi) = \mathcal{D}^{-1}$. \end{enumerate} \end{rmk}
\begin{prop} $\hat{G}$ muni de la topologie compacte-ouverte est un groupe topologique séparé : on l'appelle le dual de Pontryagin de $G$. \end{prop}
\begin{prop} $\hat{G}$ muni de la topologie compacte-ouverte est un groupe topologique séparé : on l'appelle le dual de Pontryagin de $G$. \end{prop}
\begin{prop} $C_c^{\infty}(F^*)$ est l'espace vectoriel engendré par les fonctions $\1_{a(1+\varpi^m \OO)}$, $n \geqslant 1$, $a \in F^*$. \end{prop} \begin{proof} Les fonctions $1+\varpi^m \OO$, $n \geqslant 1$, forment une base de voisinages de $1$ dans $F^*$, et sont compacts. La fin de la preuve est laissée en exercice. \end{proof}
\begin{prop} $C_c^{\infty}(F^*)$ est l'espace vectoriel engendré par les fonctions $\1_{a(1+\varpi^m \OO)}$, $n \geqslant 1$, $a \in F^*$. \end{prop} \begin{proof} Les fonctions $1+\varpi^m \OO$, $n \geqslant 1$, forment une base de voisinages de $1$ dans $F^*$, et sont compacts. La fin de la preuve est laissée en exercice. \end{proof}
\begin{rmk}~ On peut regarder l'ensemble des $\mathcal{Z}(f, \chi)$ pour $f \in C_c^{\infty}(F)$ comme une partie de $\mathbb{C}\left(q^{-s}\right)$. \\ Il s'agit en fait d'un sous-$\mathbb{C}[q^{\pm s}]$-module qui contient $1$. \\ C'est donc un idéal fractionnaire de $\mathbb{C}\left(q^{-s}\right)$ engendré par un $\frac{1}{P\left(q^{-s}\right)}$, où $P \in \mathbb{C}[X]$.\\ On peut normaliser $P$ par $P(0)=1$. On a alors $L(\chi)=\frac{1}{P\left(q^{-s}\right)}$. En effet, $\begin{aligned}[t] \mathcal{Z}\left(f\left(\varpi_{\cdot}\right), \chi\right)&=\int_{F^*} f(\varpi \chi) \chi(x) d x^\times \\ &=\chi(\varpi)^{-1} \int_{F^*} f(x) \chi(x) d x^\times \\ &=q^s \mathcal{Z}(f, \chi) . \end{aligned}$ \end{rmk}
\begin{rmk}~ On peut regarder l'ensemble des $\mathcal{Z}(f, \chi)$ pour $f \in C_c^{\infty}(F)$ comme une partie de $\mathbb{C}\left(q^{-s}\right)$. \\ Il s'agit en fait d'un sous-$\mathbb{C}[q^{\pm s}]$-module qui contient $1$. \\ C'est donc un idéal fractionnaire de $\mathbb{C}\left(q^{-s}\right)$ engendré par un $\frac{1}{P\left(q^{-s}\right)}$, où $P \in \mathbb{C}[X]$.\\ On peut normaliser $P$ par $P(0)=1$. On a alors $L(\chi)=\frac{1}{P\left(q^{-s}\right)}$. En effet, $\begin{aligned}[t] \mathcal{Z}\left(f\left(\varpi_{\cdot}\right), \chi\right)&=\int_{F^*} f(\varpi \chi) \chi(x) d x^\times \\ &=\chi(\varpi)^{-1} \int_{F^*} f(x) \chi(x) d x^\times \\ &=q^s \mathcal{Z}(f, \chi) . \end{aligned}$ \end{rmk}
\begin{rap} Si $G$ est séparable et $\exists D \subseteq G$ dénombrable et dense, alors la topologie est $\sigma$-compacte. \end{rap}
\begin{rap} Si $G$ est séparable et $\exists D \subseteq G$ dénombrable et dense, alors la topologie est $\sigma$-compacte. \end{rap}
\subsection{Fonctions zêta archimédiennes:} - On prend $F=\mathbb{R}$ ou $\mathbb{C}$. On fixe un caractère additif de $F$ : $$ \psi_{\mathbb{R}}: \mathbb{R} \rightarrow \mathbb{U}, x \mapsto e^{-2 i \pi x} $$ $$ \psi_{\mathbb{C}}=\psi_{\mathbb{R}} \circ T_{\mathbb{C}}. $$ On considère la mesure de Lebesgue $dx$ sur $\mathbb{R}$, et $d z d \bar{z}=2 d x d y $ sur $\mathbb{C}$. \begin{prop} $a \mapsto \psi_a$ réalise un isomorphisme topologique de $F$ sur son dual de Pontryagin. \end{prop} \underline{Transformation de Fourier :} On prend $\mathcal{S}(F)$ l'espace de Schwartz des fonctions $f: F \rightarrow \C$ lisses à décroissance rapide ainsi que toutes leurs dérivées. \[ \text{Si } f \in \mathcal{S}(F), \quad \hat{\hat{f}}(y)=\int_F f(x) \psi(xy) \, dx = f(-y) \quad (\text{et } \hat{\hat{f}} \in \mathcal{S}(F)). \]
\subsection{Fonctions zêta archimédiennes:} - On prend $F=\mathbb{R}$ ou $\mathbb{C}$. On fixe un caractère additif de $F$ : $$ \psi_{\mathbb{R}}: \mathbb{R} \rightarrow \mathbb{U}, x \mapsto e^{-2 i \pi x} $$ $$ \psi_{\mathbb{C}}=\psi_{\mathbb{R}} \circ T_{\mathbb{C}}. $$ On considère la mesure de Lebesgue $dx$ sur $\mathbb{R}$, et $d z d \bar{z}=2 d x d y $ sur $\mathbb{C}$. \begin{prop} $a \mapsto \psi_a$ réalise un isomorphisme topologique de $F$ sur son dual de Pontryagin. \end{prop} \underline{Transformation de Fourier :} On prend $\mathcal{S}(F)$ l'espace de Schwartz des fonctions $f: F \rightarrow \C$ lisses à décroissance rapide ainsi que toutes leurs dérivées. \[ \text{Si } f \in \mathcal{S}(F), \quad \hat{\hat{f}}(y)=\int_F f(x) \psi(xy) \, dx = f(-y) \quad (\text{et } \hat{\hat{f}} \in \mathcal{S}(F)). \]
\begin{lemme} Si $n_\chi>0$, $\mathcal{Z}(g, \chi)$ admet un prolongement holomorphe qui ne s'annule jamais. \end{lemme} \begin{proof}~ En effet, on a pour $\Re(s) > 0$ : $$ \mathcal{Z}(g, \chi) = q^{-\left(n_\psi-n_\chi\right) s} G\left(\eta, \psi_{\varpi^{n_\psi-n_\chi}}\right) . $$ \end{proof}
\begin{lemme} Si $n_\chi>0$, $\mathcal{Z}(g, \chi)$ admet un prolongement holomorphe qui ne s'annule jamais. \end{lemme} \begin{proof}~ En effet, on a pour $\Re(s) > 0$ : $$ \mathcal{Z}(g, \chi) = q^{-\left(n_\psi-n_\chi\right) s} G\left(\eta, \psi_{\varpi^{n_\psi-n_\chi}}\right) . $$ \end{proof}
\begin{lemme} Si $n_\chi=0$, c'est-à-dire si $\eta \equiv 1$, on a : $$ \mathcal{Z}(g, \chi) = \frac{\operatorname{vol}\left(\OO^{\times}, dx^{\times}\right)}{1-\chi(\varpi)} = \frac{\left(\OO^{\times}, dx^{\times}\right)}{1-q^{-s}} $$ \end{lemme}
\begin{lemme} Si $n_\chi=0$, c'est-à-dire si $\eta \equiv 1$, on a : $$ \mathcal{Z}(g, \chi) = \frac{\operatorname{vol}\left(\OO^{\times}, dx^{\times}\right)}{1-\chi(\varpi)} = \frac{\left(\OO^{\times}, dx^{\times}\right)}{1-q^{-s}} $$ \end{lemme}
\begin{Théorème} Le morphisme canonique $\left\{\begin{array}{l} G \rightarrow \widehat{\hat{G}} \\ g \mapsto ev_g \end{array}\right.$ est un isomorphisme de groupes topologiques. Ainsi, $G \simeq \hat{\hat{G}}$ naturellement. \end{Théorème}
\begin{Théorème} Le morphisme canonique $\left\{\begin{array}{l} G \rightarrow \widehat{\hat{G}} \\ g \mapsto ev_g \end{array}\right.$ est un isomorphisme de groupes topologiques. Ainsi, $G \simeq \hat{\hat{G}}$ naturellement. \end{Théorème}
\begin{lemme} Soit $\psi: F \rightarrow \mathbb{C}^*$ un caractère additif. \begin{enumerate} \item[(i)] $\psi$ est localement constant (c'est-à-dire que $\operatorname{Ker} \psi$ est ouvert). \item[(ii)] $\psi$ prend ses valeurs dans $U_{\infty} \subseteq \mathbb{U}$. \end{enumerate} \end{lemme}
\begin{lemme} Soit $\psi: F \rightarrow \mathbb{C}^*$ un caractère additif. \begin{enumerate} \item[(i)] $\psi$ est localement constant (c'est-à-dire que $\operatorname{Ker} \psi$ est ouvert). \item[(ii)] $\psi$ prend ses valeurs dans $U_{\infty} \subseteq \mathbb{U}$. \end{enumerate} \end{lemme}
\begin{proof}(\cref{thm1})~ Nous avions $\hat{f}(y) = \psi(a y) \int_{\varpi^n \OO} \psi(x y) \,dx$ $$ = \operatorname{Vol}(\varpi^n \OO, dx) \psi(a y) \1_{\varpi^{-n} \mathcal{D}^{-1}}(y) $$ Donc, $\widehat{\hat{f}} \in C_c^{\infty}(F)$. Maintenant, pour $x \in F$, on calcule : $$ \begin{aligned} \widehat{\hat{f}}(x) & = \operatorname{Vol}(\varpi^n \OO, d x) \int_F \1_{\varpi^{-n} \mathcal{D}^{-1}}(y) \psi(a y) \psi(x y) \,dy \\ & = \operatorname{Vol}(\varpi^n \OO, d x) \int_{\varpi^{-n} \mathcal{D}^{-1}} \psi((a+x) y) \,dy . \end{aligned} $$ - Si $ x=-a$, l'intégrale vaut $\operatorname{Vol}(\varpi^{-n} \mathcal{D}^{-1}, d x)$.\\ - Si $x \neq -a$, l'intégrale est nulle sauf si $\varpi^{-n} \mathcal{D}^{-1} \subseteq (a+x)^{-1} \mathcal{D}^{-1}$, c'est-à-dire si $a+x \in \varpi^n \OO$. \\ Dans ce cas, elle vaut $\operatorname{Vol}(\varpi^{-n} \mathcal{D}^{-1}, d x)$.\\ On a alors $\widehat{\hat{f}}(x) = |\varpi^n| \operatorname{Vol}(\OO, d x) \cdot |\varpi^{-n}| \operatorname{Vol}(\mathcal{D}^{-1}, d x)$. \\ Mais $\mathcal{D}^{-1} = \bigsqcup\limits_{x \in \mathcal{D}^{-1}/\OO} x + \OO$ d'où $\operatorname{Vol}(\mathcal{D}^{-1}, d x) = \sum\limits_{x \in \mathcal{D}^{-1}/\OO} \operatorname{Vol}(x + \OO, d x) = \left|\mathcal{D}^{-1} / \OO \right| \operatorname{Vol}(\OO, dx) \text { . }$\\ Donc, $\widehat{\hat{f}}(x) = \operatorname{Vol}(\OO, d x)^2 \cdot \left|\mathcal{D}^{-1} / \OO \right| = |\OO / \mathcal{D}|^{-1} \cdot \left|\mathcal{D}^{-1} / \OO\right| = 1$ car si $\mathcal{D} = \varpi^d \OO$ avec $d \geqslant 0$, alors $\mathcal{D}^{-1} = \varpi^{-d} \OO$. \\ Ainsi, $\widehat{\hat{f}}(x) = \1_{\varpi^n \OO}(a+x) = \1_{-a+\varpi^n \OO}(x) = \1_{a+\varpi^m \OO}(-x) = f(-x)$. \end{proof}
\begin{proof}(\cref{thm1})~ Nous avions $\hat{f}(y) = \psi(a y) \int_{\varpi^n \OO} \psi(x y) \,dx$ $$ = \operatorname{Vol}(\varpi^n \OO, dx) \psi(a y) \1_{\varpi^{-n} \mathcal{D}^{-1}}(y) $$ Donc, $\widehat{\hat{f}} \in C_c^{\infty}(F)$. Maintenant, pour $x \in F$, on calcule : $$ \begin{aligned} \widehat{\hat{f}}(x) & = \operatorname{Vol}(\varpi^n \OO, d x) \int_F \1_{\varpi^{-n} \mathcal{D}^{-1}}(y) \psi(a y) \psi(x y) \,dy \\ & = \operatorname{Vol}(\varpi^n \OO, d x) \int_{\varpi^{-n} \mathcal{D}^{-1}} \psi((a+x) y) \,dy . \end{aligned} $$ - Si $ x=-a$, l'intégrale vaut $\operatorname{Vol}(\varpi^{-n} \mathcal{D}^{-1}, d x)$.\\ - Si $x \neq -a$, l'intégrale est nulle sauf si $\varpi^{-n} \mathcal{D}^{-1} \subseteq (a+x)^{-1} \mathcal{D}^{-1}$, c'est-à-dire si $a+x \in \varpi^n \OO$. \\ Dans ce cas, elle vaut $\operatorname{Vol}(\varpi^{-n} \mathcal{D}^{-1}, d x)$.\\ On a alors $\widehat{\hat{f}}(x) = |\varpi^n| \operatorname{Vol}(\OO, d x) \cdot |\varpi^{-n}| \operatorname{Vol}(\mathcal{D}^{-1}, d x)$. \\ Mais $\mathcal{D}^{-1} = \bigsqcup\limits_{x \in \mathcal{D}^{-1}/\OO} x + \OO$ d'où $\operatorname{Vol}(\mathcal{D}^{-1}, d x) = \sum\limits_{x \in \mathcal{D}^{-1}/\OO} \operatorname{Vol}(x + \OO, d x) = \left|\mathcal{D}^{-1} / \OO \right| \operatorname{Vol}(\OO, dx) \text { . }$\\ Donc, $\widehat{\hat{f}}(x) = \operatorname{Vol}(\OO, d x)^2 \cdot \left|\mathcal{D}^{-1} / \OO \right| = |\OO / \mathcal{D}|^{-1} \cdot \left|\mathcal{D}^{-1} / \OO\right| = 1$ car si $\mathcal{D} = \varpi^d \OO$ avec $d \geqslant 0$, alors $\mathcal{D}^{-1} = \varpi^{-d} \OO$. \\ Ainsi, $\widehat{\hat{f}}(x) = \1_{\varpi^n \OO}(a+x) = \1_{-a+\varpi^n \OO}(x) = \1_{a+\varpi^m \OO}(-x) = f(-x)$. \end{proof}
Équation fonctionnelle: À $\chi$, est associé $\chi^{\vee}$ tel que $\chi^{\vee}(x) = |x| \chi(x)^{-1} = |x| \bar{\eta(x)} |x|^{-s}$, $$ \left. \qquad \qquad \; \; = \bar{\eta}(x) |x|^{1-s} \right. $$ En quelque sorte, $\vee: \left\{ \begin{array}{l} \eta \leftrightarrow \vec{\eta} \\ s \leftrightarrow 1-s \end{array} \right.$
Équation fonctionnelle: À $\chi$, est associé $\chi^{\vee}$ tel que $\chi^{\vee}(x) = |x| \chi(x)^{-1} = |x| \bar{\eta(x)} |x|^{-s}$, $$ \left. \qquad \qquad \; \; = \bar{\eta}(x) |x|^{1-s} \right. $$ En quelque sorte, $\vee: \left\{ \begin{array}{l} \eta \leftrightarrow \vec{\eta} \\ s \leftrightarrow 1-s \end{array} \right.$
\begin{proof}~ \begin{itemize} \item[$\rightarrow$] Si $n_\psi<n_{\chi}$, alors $\OO^{\times}= \bigsqcup\limits_{a \in \OO^{\times} / U_n} a U_n$, (où $n = n_\chi - 1 \geq n_\psi$), donc: $$ G(\chi, \psi)=\sum_{a \in \OO^{\times} / U_n} \int_{a U_n} \chi(x) \psi(x) \, dx^{\times}=\sum_{a \in \OO^{\times} / U_n} \chi(a) \int_{U_n} \chi(x) \psi(a x) \, dx^{\times}. $$ Mais $U_n=1+\varpi^n \OO$ donc $x=1+y$ avec $y \in \varpi^n \OO$ d'où $ay \in \varpi^n \OO \subseteq \varpi^{n_\psi} \OO$. \\ Ainsi, $\psi(a x)=\psi(a) \overbrace{\psi(a y)}^{=1}$, donc : $$ G(\chi, \psi)=\sum_{a \in \OO^{\times} / U_n} \psi(a) \chi(a) \underbrace{\int_{U_n} \chi(x) \, dx^{\prime}}_{= 0 \text{ car }\chi |_{U_n} \not \equiv 1} .$$ En effet $\chi |_{U_n} \not \equiv 1$ car $n_\psi<n_{\chi}$, enfin : $G(\chi, \psi)=0$. \item[$\rightarrow$] Si $n_\psi \geq n_\chi$. Puisque $n_\chi>0$, on a $n_\psi>0$ aussi. $$ \begin{aligned} |G(\chi, \psi)|_{\infty}^2 & =\int_{\OO^{\times}} \chi(x) \psi(x) \, dx^{\times} \int_{\OO^{\times}} \chi\left(y^{-1}\right) \psi(-y) \, dy^{\times} \quad\left(\begin{array}{l} \chi^{-1}=\bar{\chi} \\ \psi^{\prime}=\bar{\psi} \end{array}\right) \\ & =\int_{(\OO^{\times})^2} \chi\left(x y^{-1}\right) \psi(x-y) \, dx^{\times} \, dy^{\times} \quad (x^{\prime}=x y^{-1}) \\ & =\int_{(\OO^{\times})^2} \chi\left(x^{\prime}\right) \psi\left(\left(x^{\prime}-1\right) y\right) \, dx^{\prime \times} \, dy^{ \times} \\ & =\int_{\OO^{\times}} \chi(x) h(x) \, dx^{\times} \end{aligned} $$ $$ \begin{aligned} \text{où }h(x) :&=\int_{\OO^{\times}} \psi((x-1) y) \, dy^{\times} \\ & =\left(1-q^{-1}\right)^{-1} \int_{\OO^{\times}} \psi((x-1) y) \, dy\\ & =\left(1-q^{-1}\right)^{-1}\left(\int_{\OO} \overbrace{\psi((x-1) y)}^{= \psi_{x-1}(y)} \, dy-\int_{\varpi \OO} \psi((x-1) y) \, dy\right) \end{aligned} $$ Or $\int_{\OO} \psi_{x-1}(y) \, dy=\left\{\begin{array}{l}\operatorname{Vol}(0, dx) \text{ si } n_\psi-1 < v(x-1) \\ 0 \text{ sinon }\end{array}\right.$ \\ car $\operatorname{Cond} \left(\psi_{x-1}\right)=(x-1)^{-1} \varpi^{n_\psi} \OO=\varpi^{n_\psi-v(x-1)} \OO$. \\ Donc $h(x)= \begin{cases}\operatorname{Vol}(\OO, dx) & \text{ si } n_\psi-1<v(x-1) \\ \frac{-1}{q-1} \operatorname{Vol}(\OO, dx) & \text{ si } n_\psi-1=v(x-1) \\ 0 & \text{ si } n_\psi-1>v(x-1)\end{cases}$ \\ Ainsi $|G(\chi, \psi)|_{\infty}^2 \operatorname{Vol}(\OO, dx)^{-1}=\int_{V_{n_\psi}} \frac{-1}{q-1} \chi(x) \, dx^{\times}+\int_{U_{n_\psi}} \chi(x) \, dx^{\times} \quad \text{où }V_n: =U_{n-1} \setminus \, U_n $ $$ =\frac{-1}{q-1} \left( \int_{U_{n_\psi-1}} \chi(x) \, dx^{\times}-\int_{U_{n_\psi}} \chi(x) \, dx^{\times}\right)+\int_{U_{n_\psi}} \chi(x) \, dx^{\times} $$ Si $n_\psi=n_\chi, \int_{U_{n_\psi-1}} \chi=0$ et $|G(\chi, \psi)|_{\infty}^2 \operatorname{Vol}(\OO, dx)^{-1}=\operatorname{Vol}\left(U_{n_\psi}, dx^{\times}\right)\left(1+\frac{1}{q-1}\right)$.\\ Si $n_\psi>n_\chi, \; |G(\chi, \psi)|_{\infty}^2 \operatorname{Vol}(\OO, dx)^{-1}=\left(1-q^{-1}\right)^{-1} \operatorname{Vol}\left(U_{n_\psi}, dx^{\times}\right) -q^{-1} \operatorname{Vol}\left(U_{n_\psi - 1}, dx^{\times}\right).$ \end{itemize} \end{proof}
\begin{proof}~ \begin{itemize} \item[$\rightarrow$] Si $n_\psi<n_{\chi}$, alors $\OO^{\times}= \bigsqcup\limits_{a \in \OO^{\times} / U_n} a U_n$, (où $n = n_\chi - 1 \geq n_\psi$), donc: $$ G(\chi, \psi)=\sum_{a \in \OO^{\times} / U_n} \int_{a U_n} \chi(x) \psi(x) \, dx^{\times}=\sum_{a \in \OO^{\times} / U_n} \chi(a) \int_{U_n} \chi(x) \psi(a x) \, dx^{\times}. $$ Mais $U_n=1+\varpi^n \OO$ donc $x=1+y$ avec $y \in \varpi^n \OO$ d'où $ay \in \varpi^n \OO \subseteq \varpi^{n_\psi} \OO$. \\ Ainsi, $\psi(a x)=\psi(a) \overbrace{\psi(a y)}^{=1}$, donc : $$ G(\chi, \psi)=\sum_{a \in \OO^{\times} / U_n} \psi(a) \chi(a) \underbrace{\int_{U_n} \chi(x) \, dx^{\prime}}_{= 0 \text{ car }\chi |_{U_n} \not \equiv 1} .$$ En effet $\chi |_{U_n} \not \equiv 1$ car $n_\psi<n_{\chi}$, enfin : $G(\chi, \psi)=0$. \item[$\rightarrow$] Si $n_\psi \geq n_\chi$. Puisque $n_\chi>0$, on a $n_\psi>0$ aussi. $$ \begin{aligned} |G(\chi, \psi)|_{\infty}^2 & =\int_{\OO^{\times}} \chi(x) \psi(x) \, dx^{\times} \int_{\OO^{\times}} \chi\left(y^{-1}\right) \psi(-y) \, dy^{\times} \quad\left(\begin{array}{l} \chi^{-1}=\bar{\chi} \\ \psi^{\prime}=\bar{\psi} \end{array}\right) \\ & =\int_{(\OO^{\times})^2} \chi\left(x y^{-1}\right) \psi(x-y) \, dx^{\times} \, dy^{\times} \quad (x^{\prime}=x y^{-1}) \\ & =\int_{(\OO^{\times})^2} \chi\left(x^{\prime}\right) \psi\left(\left(x^{\prime}-1\right) y\right) \, dx^{\prime \times} \, dy^{ \times} \\ & =\int_{\OO^{\times}} \chi(x) h(x) \, dx^{\times} \end{aligned} $$ $$ \begin{aligned} \text{où }h(x) :&=\int_{\OO^{\times}} \psi((x-1) y) \, dy^{\times} \\ & =\left(1-q^{-1}\right)^{-1} \int_{\OO^{\times}} \psi((x-1) y) \, dy\\ & =\left(1-q^{-1}\right)^{-1}\left(\int_{\OO} \overbrace{\psi((x-1) y)}^{= \psi_{x-1}(y)} \, dy-\int_{\varpi \OO} \psi((x-1) y) \, dy\right) \end{aligned} $$ Or $\int_{\OO} \psi_{x-1}(y) \, dy=\left\{\begin{array}{l}\operatorname{Vol}(0, dx) \text{ si } n_\psi-1 < v(x-1) \\ 0 \text{ sinon }\end{array}\right.$ \\ car $\operatorname{Cond} \left(\psi_{x-1}\right)=(x-1)^{-1} \varpi^{n_\psi} \OO=\varpi^{n_\psi-v(x-1)} \OO$. \\ Donc $h(x)= \begin{cases}\operatorname{Vol}(\OO, dx) & \text{ si } n_\psi-1<v(x-1) \\ \frac{-1}{q-1} \operatorname{Vol}(\OO, dx) & \text{ si } n_\psi-1=v(x-1) \\ 0 & \text{ si } n_\psi-1>v(x-1)\end{cases}$ \\ Ainsi $|G(\chi, \psi)|_{\infty}^2 \operatorname{Vol}(\OO, dx)^{-1}=\int_{V_{n_\psi}} \frac{-1}{q-1} \chi(x) \, dx^{\times}+\int_{U_{n_\psi}} \chi(x) \, dx^{\times} \quad \text{où }V_n: =U_{n-1} \setminus \, U_n $ $$ =\frac{-1}{q-1} \left( \int_{U_{n_\psi-1}} \chi(x) \, dx^{\times}-\int_{U_{n_\psi}} \chi(x) \, dx^{\times}\right)+\int_{U_{n_\psi}} \chi(x) \, dx^{\times} $$ Si $n_\psi=n_\chi, \int_{U_{n_\psi-1}} \chi=0$ et $|G(\chi, \psi)|_{\infty}^2 \operatorname{Vol}(\OO, dx)^{-1}=\operatorname{Vol}\left(U_{n_\psi}, dx^{\times}\right)\left(1+\frac{1}{q-1}\right)$.\\ Si $n_\psi>n_\chi, \; |G(\chi, \psi)|_{\infty}^2 \operatorname{Vol}(\OO, dx)^{-1}=\left(1-q^{-1}\right)^{-1} \operatorname{Vol}\left(U_{n_\psi}, dx^{\times}\right) -q^{-1} \operatorname{Vol}\left(U_{n_\psi - 1}, dx^{\times}\right).$ \end{itemize} \end{proof}
\begin{prop} (Formule de Plancherel) Soient $f, g \in C_c^{\infty}(F)$. \\ On a $f \hat{g}, \hat{f g} \in C_c^{\infty}(F)$ et: $$ \int_F f(x) \hat{g}(x) d x=\int_F \hat{f}(x) g(x) d x $$ \end{prop}
\begin{prop} (Formule de Plancherel) Soient $f, g \in C_c^{\infty}(F)$. \\ On a $f \hat{g}, \hat{f g} \in C_c^{\infty}(F)$ et: $$ \int_F f(x) \hat{g}(x) d x=\int_F \hat{f}(x) g(x) d x $$ \end{prop}
\begin{rmk} On peut formuler une variante où l'on part d'un caractère $\psi_0 \neq \1$ de $F$ et où l'on en déduit les autres par translation $x \mapsto \psi_0(ax)$ pour $a \in F$. \end{rmk}
\begin{rmk} On peut formuler une variante où l'on part d'un caractère $\psi_0 \neq \1$ de $F$ et où l'on en déduit les autres par translation $x \mapsto \psi_0(ax)$ pour $a \in F$. \end{rmk}
\begin{Cor} Soit $\psi: F \rightarrow \mathbb{C}^*$ un caractère non trivial. \\ Tout caractère de $F$ est de la forme $\psi_a:=(x \mapsto \psi(a x))$ pour un unique $a \in F$. En fait, l'application $\theta:\left\{\begin{array}{l}F \rightarrow \hat{F} \text{ (dual de Pontryagin) } \\ a \mapsto \psi_a\end{array}\right.$ est un isomorphisme topologique. \end{Cor}
\begin{Cor} Soit $\psi: F \rightarrow \mathbb{C}^*$ un caractère non trivial. \\ Tout caractère de $F$ est de la forme $\psi_a:=(x \mapsto \psi(a x))$ pour un unique $a \in F$. En fait, l'application $\theta:\left\{\begin{array}{l}F \rightarrow \hat{F} \text{ (dual de Pontryagin) } \\ a \mapsto \psi_a\end{array}\right.$ est un isomorphisme topologique. \end{Cor}
\underline{Caractères multiplicatifs de $F^*$:} Pour $F=\R: $ Comme $\R^*=\{ \pm 1\} \times \R_{+}^* $,\\ Un caractère s'écrit $\chi = \left(\operatorname{sgn}\right)^{\varepsilon}\left|\cdot\right|^{s},$ avec $ \in \C, \, \varepsilon \in\{0,1\}$ et $|\cdot|$ est la valeur absolue usuelle. \\ Pour $F=\C: $ Comme $\C^*=\U \times \R_{+}^* $,\\ Un caractère s'écrit $\chi(z)=\left(\frac{z}{|z|^{1 / 2}}\right)^n|z|_{\C}^s$, avec $ n \in \Z$ et $|z|_C=z \bar{z}$ est la valeur absolue normalisée sur $\C$.\\ \underline{Facteurs $L$} La fonction $\Gamma(s)=\int_0^{+\infty} t^s e^{-t} \frac{dt}{t}$ admet un prolongement méromorphe à $\C$ holomorphe sur $\C \setminus (-\N)$ (pôles simples sur cet ensemble). \[ \Gamma_{\R}(s):=\pi^{-s / 2} \Gamma(s / 2) \qquad \Gamma_{\C}(s):=2 \cdot (2 \pi)^{-s} \Gamma(s) \] On a: \begin{aligned}[t] &L\left(\left(\frac{x}{|x|}\right)^{\varepsilon} |.|^s\right)=\Gamma_{\R}(s+\varepsilon) \\&L\left(\left(\frac{z}{|z|^{1 / 2}}\right)^n|\cdot|_{\C}^s\right)=\Gamma_{\C}\left(s+\frac{|n|}{2}\right). \end{aligned} Pour un caractère multiplicatif $\chi$ de $F^*$ et $f \in \mathcal{S}(F)$, on considère : \[ \mathcal{Z}(f, \chi)=\int_{F^*} f(x) \chi(x) \, dx^\times \quad \text{où } dx^\times=\frac{dx}{|x|}. \]
\underline{Caractères multiplicatifs de $F^*$:} Pour $F=\R: $ Comme $\R^*=\{ \pm 1\} \times \R_{+}^* $,\\ Un caractère s'écrit $\chi = \left(\operatorname{sgn}\right)^{\varepsilon}\left|\cdot\right|^{s},$ avec $ \in \C, \, \varepsilon \in\{0,1\}$ et $|\cdot|$ est la valeur absolue usuelle. \\ Pour $F=\C: $ Comme $\C^*=\U \times \R_{+}^* $,\\ Un caractère s'écrit $\chi(z)=\left(\frac{z}{|z|^{1 / 2}}\right)^n|z|_{\C}^s$, avec $ n \in \Z$ et $|z|_C=z \bar{z}$ est la valeur absolue normalisée sur $\C$.\\ \underline{Facteurs $L$} La fonction $\Gamma(s)=\int_0^{+\infty} t^s e^{-t} \frac{dt}{t}$ admet un prolongement méromorphe à $\C$ holomorphe sur $\C \setminus (-\N)$ (pôles simples sur cet ensemble). \[ \Gamma_{\R}(s):=\pi^{-s / 2} \Gamma(s / 2) \qquad \Gamma_{\C}(s):=2 \cdot (2 \pi)^{-s} \Gamma(s) \] On a: \begin{aligned}[t] &L\left(\left(\frac{x}{|x|}\right)^{\varepsilon} |.|^s\right)=\Gamma_{\R}(s+\varepsilon) \\&L\left(\left(\frac{z}{|z|^{1 / 2}}\right)^n|\cdot|_{\C}^s\right)=\Gamma_{\C}\left(s+\frac{|n|}{2}\right). \end{aligned} Pour un caractère multiplicatif $\chi$ de $F^*$ et $f \in \mathcal{S}(F)$, on considère : \[ \mathcal{Z}(f, \chi)=\int_{F^*} f(x) \chi(x) \, dx^\times \quad \text{où } dx^\times=\frac{dx}{|x|}. \]
\begin{proof}~ La topologie engendrée par les $W(K, B(\alpha, \varepsilon))$ est clairement moins fine que la compacte-ouverte. \\ Soient $K \subseteq G$ compact, $U \subseteq \U$ ouvert. Il suffit de montrer que $\forall \chi \in W(K, U)$, $\exists n$, $\exists K_i$, $\alpha_i$, $\varepsilon_i$ tels que $\chi \in \bigcap\limits\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon_i\right)\right) \subseteq W(K, U)$. \\ Pour tous $ \alpha \in U$, $\exists \varepsilon_\alpha > 0$ tel que $B\left(\alpha, \varepsilon_\alpha\right) \subseteq U$, et $U = \bigcup\limits_{\alpha \in U} B\left(\alpha, \varepsilon_\alpha\right)$. \\ Ainsi pour $\chi \in W(K,U)$, $K = \bigcup\limits\limits_{\alpha \in U} \underbrace{K \cap \chi^{-1}\left(B\left(\alpha, \varepsilon_\alpha\right)\right)}_{\text { ouvert }}$. \\ Par compacité de $K$, $\exists \alpha_1, \ldots, \alpha_n \in U $ tels que $K = \bigcup\limits_{i=1}^n K \cap \chi^{-1}\left(B\left(\alpha_i, \varepsilon_{\alpha_i}\right)\right)$. \\ Or $K$ est compact donc normal, c'est-à-dire que deux fermés de $K$ dont l'intersection est vide ont des voisinages disjoints. Ainsi, $\forall i$, $\exists K_i \subseteq K \cap \chi^{-1}\left(B\left(\alpha_i, \varepsilon_{\alpha_i}\right)\right)$ tels que $K = \bigcup\limits_{i=1}^n K_i$. \\ Mais alors : \begin{align*} \chi &\in \bigcap\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon_{\alpha_i}\right)\right) \\ &\subseteq W\left(\bigcup\limits\limits_{i=1}^n K_i, \bigcup\limits_{i=1}^n B\left(\alpha_i, \varepsilon_{\alpha_i}\right)\right) \\ &\subseteq W(K, U) \end{align*}. \end{proof} Pour $K \subseteq G$ compact, $\chi \in \hat{G}$ et $\varepsilon > 0$, on définit : $$ B_K(\chi, \varepsilon) := \{\phi \in \hat{G} \mid \forall k \in K \quad |\chi(k) - \phi(k)| < \varepsilon\} $$
\begin{proof}~ La topologie engendrée par les $W(K, B(\alpha, \varepsilon))$ est clairement moins fine que la compacte-ouverte. \\ Soient $K \subseteq G$ compact, $U \subseteq \U$ ouvert. Il suffit de montrer que $\forall \chi \in W(K, U)$, $\exists n$, $\exists K_i$, $\alpha_i$, $\varepsilon_i$ tels que $\chi \in \bigcap\limits\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon_i\right)\right) \subseteq W(K, U)$. \\ Pour tous $ \alpha \in U$, $\exists \varepsilon_\alpha > 0$ tel que $B\left(\alpha, \varepsilon_\alpha\right) \subseteq U$, et $U = \bigcup\limits_{\alpha \in U} B\left(\alpha, \varepsilon_\alpha\right)$. \\ Ainsi pour $\chi \in W(K,U)$, $K = \bigcup\limits\limits_{\alpha \in U} \underbrace{K \cap \chi^{-1}\left(B\left(\alpha, \varepsilon_\alpha\right)\right)}_{\text { ouvert }}$. \\ Par compacité de $K$, $\exists \alpha_1, \ldots, \alpha_n \in U $ tels que $K = \bigcup\limits_{i=1}^n K \cap \chi^{-1}\left(B\left(\alpha_i, \varepsilon_{\alpha_i}\right)\right)$. \\ Or $K$ est compact donc normal, c'est-à-dire que deux fermés de $K$ dont l'intersection est vide ont des voisinages disjoints. Ainsi, $\forall i$, $\exists K_i \subseteq K \cap \chi^{-1}\left(B\left(\alpha_i, \varepsilon_{\alpha_i}\right)\right)$ tels que $K = \bigcup\limits_{i=1}^n K_i$. \\ Mais alors : \begin{align*} \chi &\in \bigcap\limits_{i=1}^n W\left(K_i, B\left(\alpha_i, \varepsilon_{\alpha_i}\right)\right) \\ &\subseteq W\left(\bigcup\limits\limits_{i=1}^n K_i, \bigcup\limits_{i=1}^n B\left(\alpha_i, \varepsilon_{\alpha_i}\right)\right) \\ &\subseteq W(K, U) \end{align*}. \end{proof} Pour $K \subseteq G$ compact, $\chi \in \hat{G}$ et $\varepsilon > 0$, on définit : $$ B_K(\chi, \varepsilon) := \{\phi \in \hat{G} \mid \forall k \in K \quad |\chi(k) - \phi(k)| < \varepsilon\} $$
\begin{prop} \\ \begin{enumerate} \item[(i)] $\bigcap\limits_{i=1}^n W(K_i, U) = W\left(\bigcup\limits_{i=1}^n K_i, U\right)$ \item[(ii)] $\bigcap\limits_{i=1}^n W(K, U_i) = W\left(K, \bigcap\limits_{i=1}^n U_i\right)$ \item[(iii)] $\bigcap\limits_{i=1}^n W(K_i, U_i) \subseteq W\left(\bigcup\limits_{i=1}^n K_i, \bigcup\limits_{i=1}^n U_i\right)$ \end{enumerate} \end{prop}
\begin{prop} \\ \begin{enumerate} \item[(i)] $\bigcap\limits_{i=1}^n W(K_i, U) = W\left(\bigcup\limits_{i=1}^n K_i, U\right)$ \item[(ii)] $\bigcap\limits_{i=1}^n W(K, U_i) = W\left(K, \bigcap\limits_{i=1}^n U_i\right)$ \item[(iii)] $\bigcap\limits_{i=1}^n W(K_i, U_i) \subseteq W\left(\bigcup\limits_{i=1}^n K_i, \bigcup\limits_{i=1}^n U_i\right)$ \end{enumerate} \end{prop}
\begin{defi} Pour $f \in C_c^{\infty}(F)$, on définit sa transformée de Fourier $\hat{f} : F \rightarrow \mathbb{C}$ par: $$ \forall y \in F \quad \hat{f}(y) = \int_F f(x) \psi(x y) \,dx $$ \end{defi}
\begin{defi} Pour $f \in C_c^{\infty}(F)$, on définit sa transformée de Fourier $\hat{f} : F \rightarrow \mathbb{C}$ par: $$ \forall y \in F \quad \hat{f}(y) = \int_F f(x) \psi(x y) \,dx $$ \end{defi}
\begin{proof}~ $\exists y \in N\left(\frac{1}{r+1}\right)$ tel que $x^{r+1} = y^{r+1}$, donc $\exists \zeta \in \mathbb{U}_{r+1}$ tel que $$ \begin{aligned} & x = y \zeta, \quad y = \exp\left(\frac{2 i \pi t}{3}\right) \text{ avec } t \in \left] \frac{-1}{r+1}, \frac{1}{r+1}\right[ \\ & \zeta = \exp\left(\frac{2 i \pi}{r+1} b\right) \text{ avec } b \in \llbracket 0, r \rrbracket \\ & y \zeta = \exp\left[\frac{2 i \pi}{3}\left(t+\frac{3 b}{r+1}\right)\right] \text{ où } t+\frac{3 b}{r+1} \in \left] \frac{-1+3 b}{r+1}, \frac{1+3 b}{r+1}\right[ \\ & x = \exp\left(\frac{2 i \pi}{3} t'\right) \text{ avec } t' \in \left]-\frac{1}{r}, \frac{1}{r}\right[ \end{aligned} $$ $x=y \zeta$ est possible si $\frac{1}{r}>\frac{-1+3 b}{r+1}$, c'est-à-dire $2 r+1>3 b r$, c'est-à-dire $b=0$. Donc $\zeta = 1$ et $x=y \in N(\frac{1}{r+1})$ \end{proof} \begin{proof} (du \cref{lem})\\ Soit $v \in V$. Puisque $1_G \in V$, on a $v, v^2, \ldots, v^m \in V^{(m)}$. $$ \left\{\begin{array}{l} \chi(v) \in N(1) \\ \chi(v)^2 = \chi(v^2) \in N(1) \end{array} \Rightarrow \chi (v) \in N (\frac{1}{2}) \; \text{puis} \; \left\{\begin{array}{l} \chi(v) \in N(\frac{1}{2}) \\ \chi(v)^3 = \chi(v^3) \in N(1) \end{array} \Rightarrow \chi (v) \in N (\frac{1}{3}) $$ Par récurrence immédiate jusqu'à $r=m-1$, on obtient $\chi (v) \in N (\frac{1}{m})$ \end{proof}
\begin{proof}~ $\exists y \in N\left(\frac{1}{r+1}\right)$ tel que $x^{r+1} = y^{r+1}$, donc $\exists \zeta \in \mathbb{U}_{r+1}$ tel que $$ \begin{aligned} & x = y \zeta, \quad y = \exp\left(\frac{2 i \pi t}{3}\right) \text{ avec } t \in \left] \frac{-1}{r+1}, \frac{1}{r+1}\right[ \\ & \zeta = \exp\left(\frac{2 i \pi}{r+1} b\right) \text{ avec } b \in \llbracket 0, r \rrbracket \\ & y \zeta = \exp\left[\frac{2 i \pi}{3}\left(t+\frac{3 b}{r+1}\right)\right] \text{ où } t+\frac{3 b}{r+1} \in \left] \frac{-1+3 b}{r+1}, \frac{1+3 b}{r+1}\right[ \\ & x = \exp\left(\frac{2 i \pi}{3} t'\right) \text{ avec } t' \in \left]-\frac{1}{r}, \frac{1}{r}\right[ \end{aligned} $$ $x=y \zeta$ est possible si $\frac{1}{r}>\frac{-1+3 b}{r+1}$, c'est-à-dire $2 r+1>3 b r$, c'est-à-dire $b=0$. Donc $\zeta = 1$ et $x=y \in N(\frac{1}{r+1})$ \end{proof} \begin{proof} (du \cref{lem})\\ Soit $v \in V$. Puisque $1_G \in V$, on a $v, v^2, \ldots, v^m \in V^{(m)}$. $$ \left\{\begin{array}{l} \chi(v) \in N(1) \\ \chi(v)^2 = \chi(v^2) \in N(1) \end{array} \Rightarrow \chi (v) \in N (\frac{1}{2}) \; \text{puis} \; \left\{\begin{array}{l} \chi(v) \in N(\frac{1}{2}) \\ \chi(v)^3 = \chi(v^3) \in N(1) \end{array} \Rightarrow \chi (v) \in N (\frac{1}{3}) $$ Par récurrence immédiate jusqu'à $r=m-1$, on obtient $\chi (v) \in N (\frac{1}{m})$ \end{proof}
\begin{proof}~ $\int_F f(x)\left(\int_F g(y) \psi(x y) d y\right) d x=\int_{F \times F} f(x) g(y) \psi(x y)(d x d y)$ par Fubini, autorisé car $\int_F|f(x)|_{\infty} \int|g(y)|_{\infty} d y d x<+\infty$.\\ L'expression est symétrique en $f$ et $g$, d'où le résultat. \end{proof}
\begin{proof}~ $\int_F f(x)\left(\int_F g(y) \psi(x y) d y\right) d x=\int_{F \times F} f(x) g(y) \psi(x y)(d x d y)$ par Fubini, autorisé car $\int_F|f(x)|_{\infty} \int|g(y)|_{\infty} d y d x<+\infty$.\\ L'expression est symétrique en $f$ et $g$, d'où le résultat. \end{proof}
\begin{defi} Un caractère additif de $F$ est un morphisme continu $$ \chi: (F,+) \rightarrow (\mathbb{C}^*, \times). $$ On a donc pour tous $x, y \in F$, $$ \chi(x+y) = \chi(x) \chi(y). $$ \end{defi}
\begin{defi} Un caractère additif de $F$ est un morphisme continu $$ \chi: (F,+) \rightarrow (\mathbb{C}^*, \times). $$ On a donc pour tous $x, y \in F$, $$ \chi(x+y) = \chi(x) \chi(y). $$ \end{defi}
\subsubsection*{Construction des caractères de $F$, extension finie de $\mathbb{Q}_p$.} L'idée est de prendre une forme $\Q_{p}$-linéaire $l: F \rightarrow \Q_p$, puis de considérer $\psi \circ l$ pour $\psi$ un caractère de $\mathbb{Q}_{p}$. \\ On identifie $F$ à $\operatorname{Hom}_{\mathbb{Q}_p}\left(F, \Q_p\right)$ via la forme trace $\operatorname{Tr}_{F / \Q_p}: F \rightarrow \mathbb{Q}_{p}$.
\subsubsection*{Construction des caractères de $F$, extension finie de $\mathbb{Q}_p$.} L'idée est de prendre une forme $\Q_{p}$-linéaire $l: F \rightarrow \Q_p$, puis de considérer $\psi \circ l$ pour $\psi$ un caractère de $\mathbb{Q}_{p}$. \\ On identifie $F$ à $\operatorname{Hom}_{\mathbb{Q}_p}\left(F, \Q_p\right)$ via la forme trace $\operatorname{Tr}_{F / \Q_p}: F \rightarrow \mathbb{Q}_{p}$.
\begin{lemme} $\forall r \geq 1$. Si $x \in N\left(\frac{1}{r}\right)$ est tel que $x^{r+1} \in N(1)$, alors $x \in N\left(\frac{1}{r+1}\right)$. \end{lemme}
\begin{lemme} $\forall r \geq 1$. Si $x \in N\left(\frac{1}{r}\right)$ est tel que $x^{r+1} \in N(1)$, alors $x \in N\left(\frac{1}{r+1}\right)$. \end{lemme}
Pour $\chi=\eta |.|^s$, on a $n_\chi=n_\eta$. Considérons $g \in C_c^\infty(F)$ définie par : $$ g(x)= \begin{cases} \psi(x) & \text{ si } x \in \varpi^{n_\psi-n_\chi} \ . \\ 0 & \text{ sinon } \end{cases} $$ $\rightarrow$ Si $n_\chi>0$ : $$ \begin{aligned} &\mathcal{Z}(g, \chi)=\int_{F^{\times}} g(x) \chi(x) \, dx^{\times} \\ & =\int_{F^{\times} \cap \varpi^{n_\psi-n_\chi} \OO} \psi(x) \chi(x) \, dx^{\times} \\ & =\sum_{n \geqslant n_\psi-n_\chi} \chi ( \varpi )^n \int_{\OO^{\times}} \psi(\varpi^n x) \overbrace{\chi(x)}^{\eta(x)} \, dx^{\times} \\ & =\sum_{n \geqslant n_\psi-n_\chi} \chi(\varpi)^n G\left(\chi, \psi_{\varpi^n}\right) \end{aligned} $$ Or $ G\left(\chi, \psi_{\varpi^n}\right)=0$ si $n_\chi>n_\psi-n$ c'est-à-dire si $n>n_\psi-n_\chi$. \\ Ainsi, $\mathcal{Z}(g, \chi)=q^{\left(n_\chi-n_\psi\right) s} \underbrace{G\left(\chi, \psi_{\varpi^{n_\psi-n_\chi}}\right)}_{| \cdot |^2 \neq 0}$ d'où :
Pour $\chi=\eta |.|^s$, on a $n_\chi=n_\eta$. Considérons $g \in C_c^\infty(F)$ définie par : $$ g(x)= \begin{cases} \psi(x) & \text{ si } x \in \varpi^{n_\psi-n_\chi} \ . \\ 0 & \text{ sinon } \end{cases} $$ $\rightarrow$ Si $n_\chi>0$ : $$ \begin{aligned} &\mathcal{Z}(g, \chi)=\int_{F^{\times}} g(x) \chi(x) \, dx^{\times} \\ & =\int_{F^{\times} \cap \varpi^{n_\psi-n_\chi} \OO} \psi(x) \chi(x) \, dx^{\times} \\ & =\sum_{n \geqslant n_\psi-n_\chi} \chi ( \varpi )^n \int_{\OO^{\times}} \psi(\varpi^n x) \overbrace{\chi(x)}^{\eta(x)} \, dx^{\times} \\ & =\sum_{n \geqslant n_\psi-n_\chi} \chi(\varpi)^n G\left(\chi, \psi_{\varpi^n}\right) \end{aligned} $$ Or $ G\left(\chi, \psi_{\varpi^n}\right)=0$ si $n_\chi>n_\psi-n$ c'est-à-dire si $n>n_\psi-n_\chi$. \\ Ainsi, $\mathcal{Z}(g, \chi)=q^{\left(n_\chi-n_\psi\right) s} \underbrace{G\left(\chi, \psi_{\varpi^{n_\psi-n_\chi}}\right)}_{| \cdot |^2 \neq 0}$ d'où :
\begin{rmk} \begin{enumerate} \item[1)] $\1$ est le caractère trivial : $\forall g \in G \; , \; \1(g) = 1$, alors on a $B_K(\1, \varepsilon) = W(K, B(1, \varepsilon))$ c'est donc un ouvert, et $B_K(\chi, \varepsilon) = \chi \cdot B_K(\1, \varepsilon)$ l'est aussi. \item[2)] Soient $m \geq 1$ un entier et $V$ une partie de $G$. On note $V^{(m)}$ l'image de $V^m$ par $\left\{\begin{array}{l}G^m \rightarrow G \\ (g_1, \ldots, g_m) \mapsto g_1 \dots g_m\end{array}\right.$ \end{enumerate} Pour $\varepsilon > 0$, on note $N(\varepsilon)$ l'image de $]\frac{-\varepsilon}{3}, \frac{\varepsilon}{3}[$ par $\left\{\begin{array}{l}\mathbb{R} \longrightarrow \mathbb{U} \\ t \mapsto e^{2 i \pi t}\end{array}\right.$. \end{rmk}
\begin{rmk} \begin{enumerate} \item[1)] $\1$ est le caractère trivial : $\forall g \in G \; , \; \1(g) = 1$, alors on a $B_K(\1, \varepsilon) = W(K, B(1, \varepsilon))$ c'est donc un ouvert, et $B_K(\chi, \varepsilon) = \chi \cdot B_K(\1, \varepsilon)$ l'est aussi. \item[2)] Soient $m \geq 1$ un entier et $V$ une partie de $G$. On note $V^{(m)}$ l'image de $V^m$ par $\left\{\begin{array}{l}G^m \rightarrow G \\ (g_1, \ldots, g_m) \mapsto g_1 \dots g_m\end{array}\right.$ \end{enumerate} Pour $\varepsilon > 0$, on note $N(\varepsilon)$ l'image de $]\frac{-\varepsilon}{3}, \frac{\varepsilon}{3}[$ par $\left\{\begin{array}{l}\mathbb{R} \longrightarrow \mathbb{U} \\ t \mapsto e^{2 i \pi t}\end{array}\right.$. \end{rmk}
\begin{defi} La différente inverse de $F / \mathbb{Q}_{p}$ est définie comme $$ \mathcal{D}_{F / \mathbb{Q}_{p}}^{-1}:=\left\{x \in F \mid \forall y \in \OO \quad \operatorname{Tr}_{F / \mathbb{Q}_p}(x y) \in \mathbb{Z}_{p}\right\}. $$ C'est un idéal fractionnaire de $F$ qui contient $\OO$. La différente de $F / \mathbb{Q}_p$ est définie comme $\mathcal{D}_{F / \mathbb{Q}_p}=\left(\mathcal{D}_{F / \mathbb{Q}_{p}}^{-1}\right)^{-1}$ au sens des inverses d'idéaux. C'est un idéal de $\OO$. \end{defi}
\begin{defi} La différente inverse de $F / \mathbb{Q}_{p}$ est définie comme $$ \mathcal{D}_{F / \mathbb{Q}_{p}}^{-1}:=\left\{x \in F \mid \forall y \in \OO \quad \operatorname{Tr}_{F / \mathbb{Q}_p}(x y) \in \mathbb{Z}_{p}\right\}. $$ C'est un idéal fractionnaire de $F$ qui contient $\OO$. La différente de $F / \mathbb{Q}_p$ est définie comme $\mathcal{D}_{F / \mathbb{Q}_p}=\left(\mathcal{D}_{F / \mathbb{Q}_{p}}^{-1}\right)^{-1}$ au sens des inverses d'idéaux. C'est un idéal de $\OO$. \end{defi}
\begin{proof}~ Si $n_\chi>0$, on a pour $\Re(s)<1$ : \\ $\mathcal{Z}\left(\hat{g}, \chi^\vee\right)=\int_{F^*} \hat{g}(x) \chi^\vee(x) dx^\times=| \OO / \mathcal{D} |^{-1/2} q^{n_\chi-n_\psi} \underbrace{\int_{-U_{n_\chi}} \chi^\vee(x) dx^\times}_{ = \chi^\vee (-1) \operatorname{Vol}(U_{n_\chi},dx^\times)} $. \\ \ \\ Or $\eta(-1)= \pm 1$ et $\chi^\vee(-1)=\bar{\eta(-1)}=\eta(-1) = \chi(-1)$, et: $\operatorname{Vol}\left(U_{n_\chi}, dx^\times\right)=\left((q-1) q^{n_\chi-1}\right)^{-1} \operatorname{Vol}\left(\OO^\times, dx^\times\right)$.\\ D'où $\mathcal{Z}\left(\hat{g}, \chi^\vee\right)=| \OO / \mathcal{D} |^{-1} q^{-n_\psi } \frac{\chi(-1)}{1-q^{-1}}$.\\ Si $n_\chi=0 :\begin{aligned}[t] \mathcal{Z}\left(\hat{g}, \chi^\vee\right)&=\int_{F^*}| \OO / \mathcal{D} |^{-1/2} q^{-n_\psi} \1_{\OO}(x) \chi^\vee(x) dx^\times \\ & =| \OO / \mathcal{D} |^{-1/2} q^{-n_\psi} \sum_{k=0}^{+\infty} q^{-k(1-s)} \operatorname{Vol}\left(\O\OO^\times, dx^\times\right)=| \OO / \mathcal{D} |^{-1} \frac{q^{-n_\psi}}{1-q^{s-1}} \\ & \end{aligned} \end{proof} \subsection{Facteurs $L$, $\gamma $ et $\varepsilon$} \begin{defi} $g$ étant fixée comme avant, on définit le facteur gamma par: $$ \gamma(\chi, \psi)=\frac{\mathcal{Z}(\hat{g}, \chi^\vee)}{\mathcal{Z}(g, \chi)} $$ \(\gamma\) est une fonction méromorphe de $s$. \end{defi} \begin{Théorème}~ \begin{enumerate} \item[(i)] $\gamma(\chi, \psi)$ est une fonction méromorphe de $s$, égale à: \\ $\begin{aligned}[t] & \rightarrow | \OO / \mathcal{D} |^{-1} q^{-n_\psi} \frac{1-q^{-s}}{1-q^{s-1}} \quad &\text { si } n_\chi=0 \\ & \rightarrow | \OO / \mathcal{D} |^{-1} \frac{q^{-n_\psi}}{1-q^{-1}} \eta(-1) q^{\left(n_\psi-n_\chi\right) s} G(\eta, \psi_{\varpi^{n_\psi-n_\chi}})^{-1} &\text { si } n_\chi > 0 \end{aligned}$\\ Pour $\psi = \psi_p \circ \operatorname{Tr}_{F/\Q_p}$ pour lequel $dx$ est autoduale, son conducteur est $\mathcal{D}^{-1} = \varpi^d \OO$ et cette formule vaut 1. \item[(ii)] On a l'équation fonctionnelle: $$ \mathcal{Z}(\hat{f}, \chi^\vee)=\gamma(\chi, \psi) \mathcal{Z}(f, \chi) $$ \end{enumerate} \end{Théorème} \begin{defi} On définit aussi le facteur \(L\) par: \(L(\chi):=\left\{\begin{array}{lc}1 &\text {si $\chi$ non ramifié, ie si } n_\chi > 0 \\ \frac{1}{1-\chi(\varpi)}=\frac{1}{1-q^{-s}} &\text{sinon}\end{array}\right.\) Et le facteur epsilon : \(\varepsilon(\chi, \psi)=\gamma(\chi, \psi) \frac{L(\chi)}{L(\chi^\vee)}\). \end{defi} \begin{prop} La fonction \(\varepsilon\) est holomorphe en $s$ et à valeurs dans \(\mathbb{C}^*\). \end{prop} \begin{EX} \begin{enumerate} \item Calcul non ramifié : \(\OO=\mathcal{D}, n_\chi=n_\psi=0\) : $$ \varepsilon(\chi, \psi)=|\OO / \mathcal{D}|^{-1} q^{-n_\psi}=1 $$ \item Pour \(s=1 / 2\) (le centre de la symétrie \(s \leftrightarrow 1-s\)) et \(\psi\) quelconque de conducteur \(\OO\) (et \(dx\) la mesure autoduale par rapport à \(\psi\)), on a: $$ \left|\varepsilon(\eta |.|^{1 / 2}, \psi)\right|=1 $$ \end{enumerate} \end{EX} \begin{Théorème} Soient \(\eta\) un caractère de \(\OO^\times, \chi=\eta |.|^s\) et \(f \in C_c^{\infty}(F)\). \(\frac{\mathcal{Z}(f, \chi)}{L(\chi)}\) est holomorphe en $s$. \\ Mieux: c'est un polynôme en \(q^{ \pm s}\) vérifiant l'équation fonctionnelle: \(\frac{\mathcal{Z}(\hat{f}, \chi^\vee)}{L(\chi^\vee)}=\varepsilon(\chi, \psi) \frac{\mathcal{Z}(f, \chi)}{L(\chi)}\). \end{Théorème}
\begin{proof}~ Si $n_\chi>0$, on a pour $\Re(s)<1$ : \\ $\mathcal{Z}\left(\hat{g}, \chi^\vee\right)=\int_{F^*} \hat{g}(x) \chi^\vee(x) dx^\times=| \OO / \mathcal{D} |^{-1/2} q^{n_\chi-n_\psi} \underbrace{\int_{-U_{n_\chi}} \chi^\vee(x) dx^\times}_{ = \chi^\vee (-1) \operatorname{Vol}(U_{n_\chi},dx^\times)} $. \\ \ \\ Or $\eta(-1)= \pm 1$ et $\chi^\vee(-1)=\bar{\eta(-1)}=\eta(-1) = \chi(-1)$, et: $\operatorname{Vol}\left(U_{n_\chi}, dx^\times\right)=\left((q-1) q^{n_\chi-1}\right)^{-1} \operatorname{Vol}\left(\OO^\times, dx^\times\right)$.\\ D'où $\mathcal{Z}\left(\hat{g}, \chi^\vee\right)=| \OO / \mathcal{D} |^{-1} q^{-n_\psi } \frac{\chi(-1)}{1-q^{-1}}$.\\ Si $n_\chi=0 :\begin{aligned}[t] \mathcal{Z}\left(\hat{g}, \chi^\vee\right)&=\int_{F^*}| \OO / \mathcal{D} |^{-1/2} q^{-n_\psi} \1_{\OO}(x) \chi^\vee(x) dx^\times \\ & =| \OO / \mathcal{D} |^{-1/2} q^{-n_\psi} \sum_{k=0}^{+\infty} q^{-k(1-s)} \operatorname{Vol}\left(\O\OO^\times, dx^\times\right)=| \OO / \mathcal{D} |^{-1} \frac{q^{-n_\psi}}{1-q^{s-1}} \\ & \end{aligned} \end{proof} \subsection{Facteurs $L$, $\gamma $ et $\varepsilon$} \begin{defi} $g$ étant fixée comme avant, on définit le facteur gamma par: $$ \gamma(\chi, \psi)=\frac{\mathcal{Z}(\hat{g}, \chi^\vee)}{\mathcal{Z}(g, \chi)} $$ \(\gamma\) est une fonction méromorphe de $s$. \end{defi} \begin{Théorème}~ \begin{enumerate} \item[(i)] $\gamma(\chi, \psi)$ est une fonction méromorphe de $s$, égale à: \\ $\begin{aligned}[t] & \rightarrow | \OO / \mathcal{D} |^{-1} q^{-n_\psi} \frac{1-q^{-s}}{1-q^{s-1}} \quad &\text { si } n_\chi=0 \\ & \rightarrow | \OO / \mathcal{D} |^{-1} \frac{q^{-n_\psi}}{1-q^{-1}} \eta(-1) q^{\left(n_\psi-n_\chi\right) s} G(\eta, \psi_{\varpi^{n_\psi-n_\chi}})^{-1} &\text { si } n_\chi > 0 \end{aligned}$\\ Pour $\psi = \psi_p \circ \operatorname{Tr}_{F/\Q_p}$ pour lequel $dx$ est autoduale, son conducteur est $\mathcal{D}^{-1} = \varpi^d \OO$ et cette formule vaut 1. \item[(ii)] On a l'équation fonctionnelle: $$ \mathcal{Z}(\hat{f}, \chi^\vee)=\gamma(\chi, \psi) \mathcal{Z}(f, \chi) $$ \end{enumerate} \end{Théorème} \begin{defi} On définit aussi le facteur \(L\) par: \(L(\chi):=\left\{\begin{array}{lc}1 &\text {si $\chi$ non ramifié, ie si } n_\chi > 0 \\ \frac{1}{1-\chi(\varpi)}=\frac{1}{1-q^{-s}} &\text{sinon}\end{array}\right.\) Et le facteur epsilon : \(\varepsilon(\chi, \psi)=\gamma(\chi, \psi) \frac{L(\chi)}{L(\chi^\vee)}\). \end{defi} \begin{prop} La fonction \(\varepsilon\) est holomorphe en $s$ et à valeurs dans \(\mathbb{C}^*\). \end{prop} \begin{EX} \begin{enumerate} \item Calcul non ramifié : \(\OO=\mathcal{D}, n_\chi=n_\psi=0\) : $$ \varepsilon(\chi, \psi)=|\OO / \mathcal{D}|^{-1} q^{-n_\psi}=1 $$ \item Pour \(s=1 / 2\) (le centre de la symétrie \(s \leftrightarrow 1-s\)) et \(\psi\) quelconque de conducteur \(\OO\) (et \(dx\) la mesure autoduale par rapport à \(\psi\)), on a: $$ \left|\varepsilon(\eta |.|^{1 / 2}, \psi)\right|=1 $$ \end{enumerate} \end{EX} \begin{Théorème} Soient \(\eta\) un caractère de \(\OO^\times, \chi=\eta |.|^s\) et \(f \in C_c^{\infty}(F)\). \(\frac{\mathcal{Z}(f, \chi)}{L(\chi)}\) est holomorphe en $s$. \\ Mieux: c'est un polynôme en \(q^{ \pm s}\) vérifiant l'équation fonctionnelle: \(\frac{\mathcal{Z}(\hat{f}, \chi^\vee)}{L(\chi^\vee)}=\varepsilon(\chi, \psi) \frac{\mathcal{Z}(f, \chi)}{L(\chi)}\). \end{Théorème}
\begin{proof}~ \begin{enumerate} \item[(i)] Soit $U \ni 1$ un ouvert de $\C^*$.\\ $\eta^{-1}(U)$ est un ouvert contenant $1$. Il existe donc $n \geq 1$ tel que $1+\varpi^n \OO \subseteq \eta^{-1}(U)$. \\ Si $U$ est assez petit, on sait que le seul sous-groupe de $C^*$ inclus dans $U$ est $\{1\}$. \\ Pour un tel $U$, $\eta\left(1+\varpi^n \OO\right)$ est donc trivial, d'où $1+\varpi^n \OO \subseteq \operatorname{ker} \eta$. \item[(ii)] Ainsi, $\eta$ se factorise par le quotient fini $\OO^{\times} / (1+\varpi^n \OO)$, d'où Im $\eta$ est finie. \end{enumerate} \end{proof}
\begin{proof}~ \begin{enumerate} \item[(i)] Soit $U \ni 1$ un ouvert de $\C^*$.\\ $\eta^{-1}(U)$ est un ouvert contenant $1$. Il existe donc $n \geq 1$ tel que $1+\varpi^n \OO \subseteq \eta^{-1}(U)$. \\ Si $U$ est assez petit, on sait que le seul sous-groupe de $C^*$ inclus dans $U$ est $\{1\}$. \\ Pour un tel $U$, $\eta\left(1+\varpi^n \OO\right)$ est donc trivial, d'où $1+\varpi^n \OO \subseteq \operatorname{ker} \eta$. \item[(ii)] Ainsi, $\eta$ se factorise par le quotient fini $\OO^{\times} / (1+\varpi^n \OO)$, d'où Im $\eta$ est finie. \end{enumerate} \end{proof}
\begin{Théorème} Soit $G$ un groupe topologique (abélien) localement compact. \begin{enumerate} \item[(i)] Si $G$ est discret, alors $\hat{G}$ est compact. \item[(ii)] Si $G$ est compact, alors $\hat{G}$ est discret. \item[(iii)] $\hat{G}$ est un groupe topologique localement compact. \end{enumerate} \end{Théorème}
\begin{Théorème} Soit $G$ un groupe topologique (abélien) localement compact. \begin{enumerate} \item[(i)] Si $G$ est discret, alors $\hat{G}$ est compact. \item[(ii)] Si $G$ est compact, alors $\hat{G}$ est discret. \item[(iii)] $\hat{G}$ est un groupe topologique localement compact. \end{enumerate} \end{Théorème}
\begin{proof}~ Soit $\left(e_i\right)_{1 \leq i \leq n}$ une $\Q_p$-base de $F$, on note $\left(e_i^*\right)$ la base duale, de sorte que : \\ $\forall x \in F$, $x=\sum\limits_{i=1}^n e_i^*(x) e_i$. \\ Soit $\psi$ un caractère additif de $F$. \\ Pour chaque $i$, $\left\{ \begin{array}{l} \mathbb{Q}_{p} \longrightarrow \mathbb{C}^* \\ y \mapsto \psi(y e_i) \end{array} \right.$ est un caractère de $\mathbb{Q}_{p}$, signifiant qu'il existe un unique $a_i \in \Q_{p}$ tel que :\\ $\forall y \in \mathbb{Q}_{p}$, $\psi(y e_i)=\psi_{p}(a_i y)$.\\ Alors $\psi(x)=\prod\limits_{i=1}^n \psi(e_i^*(x) e_i)=\prod\limits_{i=1}^n \psi_{p}(a_i e_i^*(x))=\psi_p(l(x))$, où $l=\sum\limits_{i=1}^n a_i e_i^*$ est une forme $\Q_{p}$-linéaire de $F$.\\ \ \\ Il nous reste à montrer qu'il existe un unique $a \in F$ tel que $\forall x \in F$, $l(x)=\operatorname{Tr}_{F/Q_p}(a x)$.\\ \underline{Unicité} : Si $\forall x \in F$, $\operatorname{Tr}_{F/Q_p}(a x)=\operatorname{Tr}_{F/\Q_p}(bx)$, ie $\operatorname{Tr}_{F/\Q_p}((a-b)x)=0$, donc $a-b=0$. \\ \underline{Existence} : $\left\{ \begin{array}{l} F \longrightarrow \operatorname{Hom}_{\Q_{p}}(F, \Q_{p}) \\ a \longmapsto \operatorname{Tr}_{F/\Q_{p}}(a \, .) \end{array} \right.$ est linéaire, et injective d'après l'argument d'unicité ci-dessus. Par un argument de dimension, c'est un isomorphisme (en particulier surjectif). \end{proof}
\begin{proof}~ Soit $\left(e_i\right)_{1 \leq i \leq n}$ une $\Q_p$-base de $F$, on note $\left(e_i^*\right)$ la base duale, de sorte que : \\ $\forall x \in F$, $x=\sum\limits_{i=1}^n e_i^*(x) e_i$. \\ Soit $\psi$ un caractère additif de $F$. \\ Pour chaque $i$, $\left\{ \begin{array}{l} \mathbb{Q}_{p} \longrightarrow \mathbb{C}^* \\ y \mapsto \psi(y e_i) \end{array} \right.$ est un caractère de $\mathbb{Q}_{p}$, signifiant qu'il existe un unique $a_i \in \Q_{p}$ tel que :\\ $\forall y \in \mathbb{Q}_{p}$, $\psi(y e_i)=\psi_{p}(a_i y)$.\\ Alors $\psi(x)=\prod\limits_{i=1}^n \psi(e_i^*(x) e_i)=\prod\limits_{i=1}^n \psi_{p}(a_i e_i^*(x))=\psi_p(l(x))$, où $l=\sum\limits_{i=1}^n a_i e_i^*$ est une forme $\Q_{p}$-linéaire de $F$.\\ \ \\ Il nous reste à montrer qu'il existe un unique $a \in F$ tel que $\forall x \in F$, $l(x)=\operatorname{Tr}_{F/Q_p}(a x)$.\\ \underline{Unicité} : Si $\forall x \in F$, $\operatorname{Tr}_{F/Q_p}(a x)=\operatorname{Tr}_{F/\Q_p}(bx)$, ie $\operatorname{Tr}_{F/\Q_p}((a-b)x)=0$, donc $a-b=0$. \\ \underline{Existence} : $\left\{ \begin{array}{l} F \longrightarrow \operatorname{Hom}_{\Q_{p}}(F, \Q_{p}) \\ a \longmapsto \operatorname{Tr}_{F/\Q_{p}}(a \, .) \end{array} \right.$ est linéaire, et injective d'après l'argument d'unicité ci-dessus. Par un argument de dimension, c'est un isomorphisme (en particulier surjectif). \end{proof}
\begin{prop} L'espace vectoriel complexe $C_c^{\infty}(F)$ est engendré par les fonctions caractéristiques des ensembles $a+\varpi^n \OO$, $n \in \mathbb{Z}$, $a \in F$. \end{prop}
\begin{prop} L'espace vectoriel complexe $C_c^{\infty}(F)$ est engendré par les fonctions caractéristiques des ensembles $a+\varpi^n \OO$, $n \in \mathbb{Z}$, $a \in F$. \end{prop}
\begin{lemme} \label{lem2} Pour $\phi$ caractère non trivial et $\varpi^m \OO$ son conducteur, on a: $$\int_{\varpi^n \OO} \phi(x) \,dx = \left\{\begin{array}{ll} \operatorname{Vol}(\varpi^n \OO, dx) & \text{si }n \geq m\\ 0 & \text{si }n < m \end{array} \right. $$ $$\text{Si } \phi = 1 \text{ alors } \int_{\varpi^n \OO} \phi(x) \,dx = \operatorname{Vol}(\varpi^n \OO, dx)$$ \end{lemme}
\begin{lemme} \label{lem2} Pour $\phi$ caractère non trivial et $\varpi^m \OO$ son conducteur, on a: $$\int_{\varpi^n \OO} \phi(x) \,dx = \left\{\begin{array}{ll} \operatorname{Vol}(\varpi^n \OO, dx) & \text{si }n \geq m\\ 0 & \text{si }n < m \end{array} \right. $$ $$\text{Si } \phi = 1 \text{ alors } \int_{\varpi^n \OO} \phi(x) \,dx = \operatorname{Vol}(\varpi^n \OO, dx)$$ \end{lemme}
\begin{prop} Tout caractère additif $\psi: F \rightarrow \mathbb{C}^*$ est de la forme $$ \psi: \left\{ \begin{array}{l} F \rightarrow \mathbb{C}^* \\ x \mapsto \psi_{p}\left(\operatorname{Tr}_{F/Q_{p}}(a x)\right) \end{array} \right. \text{pour un unique $a \in F$.} $$ $\psi$ est trivial ssi $a=0$. Si $a \neq 0$, alors on a $ \operatorname{Cond}(\psi)=a^{-1} \mathcal{D}_{F/\Q_p}^{-1}$. \end{prop}
\begin{prop} Tout caractère additif $\psi: F \rightarrow \mathbb{C}^*$ est de la forme $$ \psi: \left\{ \begin{array}{l} F \rightarrow \mathbb{C}^* \\ x \mapsto \psi_{p}\left(\operatorname{Tr}_{F/Q_{p}}(a x)\right) \end{array} \right. \text{pour un unique $a \in F$.} $$ $\psi$ est trivial ssi $a=0$. Si $a \neq 0$, alors on a $ \operatorname{Cond}(\psi)=a^{-1} \mathcal{D}_{F/\Q_p}^{-1}$. \end{prop}

Dataset Card for "LatexCorrection"

More Information needed

Downloads last month
2
Edit dataset card