Height pairings, exceptional zeros and Rubin's formula: the multiplicative group (Q662441)

The author proves a formula ``à la'' Rubin which expresses the leading coefficients of various \(p\)-adic \(L\)-functions in the presence of an exceptional zero in terms of Nekovář's \(p\)-adic height pairings on extended Selmer groups. Actually, Nekovář's machinery is only used in the following classical setting: fix an odd rational prime \(p,\) a number field \(K\), the ring of integers \({\mathcal O}\) of a finite extension of \({\mathbb Q}_p\), a non-trivial Dirichlet character \(\chi: G_K \to {\mathcal O}^\times\) which has order prime to \(p\) and satisfies \(\chi ({\mathcal P}) = 1\) for a \(K\)-prime \({\mathcal P} \mid p\) and \(\chi ({\mathcal P}') \neq 1\) for all other \({\mathcal P}' \mid p\). Define \(T = {\mathcal O} (1) \otimes \chi^{-1}\) and \(T^\ast = {\mathcal O} (\chi)\), viewed as representations of \(G_K\). Then Nekovář's extended Selmer groups are \(\widetilde H^1_f (K, T^\ast) \simeq \bigoplus_{{\mathcal P} | p} H^0 (K_{\mathcal P}, {\mathcal O}(\chi))\) and \(\widetilde H^1_f (K, T) = ({\mathcal O}_L[1/p]^\times)^{\chi},\) where \(L\) is the field cut out by \(\chi.\) Nekovář's height pairing \(\widetilde H^1_f (K, T) \times \widetilde H^1_f (K, \widetilde T) \to {\mathcal O}\) is related to the classical Poitou-Tate pairing by the formula \(\langle x, y \rangle_{\text{Nek}} = \langle \beta^1_\chi (x), y \rangle_{PT},\) where \(\beta^1_\chi: \widetilde H^1_f (K, T) \to \widetilde H^2_f (K, T)\) is the Bockstein morphism. The author studies three cases: 1) \(K = {\mathbb Q}\) and \(\chi\) is even: let \(c^\chi_1 = e_\chi N_{{\mathbb Q}(\mu_f)/L} (\xi_f - 1) \in H^1 ({\mathbb Q}, T)\) be the usual ``tame \(\chi\)-cyclotomic unit'' of \(L\) (where \(f\) is the conductor of \(\chi\)), and let \(z^\chi_0 \in H^1 ({\mathbb Q}, T)\) be the ``cyclotomic \(p\)-unit'' constructed by \textit{D. Solomon} [Invent. Math. 109, No. 2, 329--350 (1992; Zbl 0772.11043)]. Fix an embedding \(\overline{\mathbb Q} \hookrightarrow \overline{\mathbb Q}_p\) which determines a place \(v_0\) of \(L\) above \(p.\) The author computes \(\langle c^\chi_1, \alpha \rangle_{\text{Nek}}\) to be \(v_0 (e^\chi_0) \ldotp \alpha (v_0),\) where \(\alpha (v_0)\) is the image of \(\alpha\) under the composite map \(\widetilde H^1_f ({\mathbb Q}, T^\ast) \overset{\sim}{\rightarrow} H^0 ({\mathbb Q}_p, {\mathcal O}(\chi)) \overset{\sim}{\rightarrow} \bigl( \bigoplus_{v | p}\, {\mathcal O} \ldotp v \bigl)^{\chi^{-1}}\;\overset{pr_{\chi^{-1}}} {\longrightarrow}{\mathcal O}.\) He deduces from this his formula ``à la'' Rubin expressing \(\langle c_1^\chi, \alpha \rangle_{Nek}\) as a certain derivative \({ d \over ds} L_{\xi, \Phi} \;\bigl( \rho^s_{\text{cyc}} \bigl) |_{s=0}.\) Here \(\rho_{\text{cyc}}\) is the cyclotomic character; \(\xi = \{ \xi_n \} \in \lim_\leftarrow H^1 ({\mathbb Q}_n, T)\) is the collection of ``wild \(\chi\)-cyclotomic units'' defined by \(\xi_n = N_{{\mathbb Q}(\mu_p f_n)/L_n} (\xi_{p f_n} - 1)\) \((L_n = L\;{\mathbb Q}_n\) and \(f_n\) is its conductor); the definition of the ``\(p\)-adic \(L\)-function'' \(L_{\xi, \Phi},\) being too technical to be recalled we'll be content to state that, via Coleman's theory for an appropriate \(\alpha,\) the author can associate to \(\alpha (v_0)\) an element \(\text{col}^\chi_0 \in \widetilde H^1_f ({\mathbb Q}, T^\ast)\) such that the above formula takes a more familiar form: \(\langle c^\chi_1, \text{col}^\chi_0 \rangle_{\text{Nek}} = \widetilde L'_p (1, \chi),\) where \(\widetilde L_p (s, \chi)\) is the ``imprimitive \(p\)-adic \(L\)-function'' defined by \(\widetilde L_p (s, \chi) = \rho_{\text{cyc}}^{1-s} (\widetilde {\mathcal L}_\chi),\) \(\widetilde {\mathcal L}_\chi = { \gamma - 1 \over {{1 \over p} \log_p \rho_{\text{cyc}}(\gamma)}} {\mathcal L}_\chi\) (where \(\gamma\) is a topological generator of \(G({\mathbb Q}_\infty / {\mathbb Q})),\) \(L_p (s, \chi) = \rho^{1-s}_{\text{cyc}} ({\mathcal L}_\chi)\). 2) \(K = {\mathbb Q}\) and \(\chi\) is odd: take a prime \({\mathcal P}\) of \(L\) above \(p\) and fix an embedding \(\iota_{\mathcal P}: L \hookrightarrow L_{\mathcal P} = {\mathbb Q}_p.\) Take a generator \(z \in \bigl( {\mathcal O}_L [1/p]^\chi \bigl)^\times = \widetilde H^1_f ({\mathbb Q}, T)\) of \(e_\chi \;{\mathcal P}^h,\) where \(h\) is the class number of \(L\). For \(z_0 = {1 \over h} \ldotp z,\) Greenberg's invariant is defined by \({\mathcal L} = log_p\;\iota_{\mathcal P} (z_0) \in \text{Frac} ({\mathcal O}).\) The author proves a formula ``à la'' Rubin along the same lines as in case 1): \(\langle z, \chi (\theta_f) \rangle_{Nek} = - {\mathcal L} \ldotp L(0, \chi^{-1}),\) where \(\theta_f\) is the ``tame Stickelberger element'' which satisfies \(\chi (\theta_f) = B_{1, \chi^{-1}}\). Note that by Ferrero-Greenberg and Gross-Koblitz, \(- {\mathcal L} \ldotp L (0, \chi^{-1}) = {d \over ds} \;L_p (s, \chi^{-1} \omega) |_{s=0}.\) 3) \(K\) is an imaginary quadratic field in which \(p\) splits: \(p = {\mathcal P} {\mathcal P}^\ast, {\mathcal P} \not= {\mathcal P}^\ast\). For any Dirichlet character \(\chi\) of \(G_K\) of order prime to \(p,\) such that \(\chi ({\mathcal P}) = 1,\) the author states that, replacing cyclotomic units by elliptic units, and the results of D. Solomon op. cit. by that of \textit{W. Bley} [J. Reine Angew. Math. 577, 117--146 (2004; Zbl 1056.11035)], he can prove a formula analogous to case 1), namely: \(\langle \varepsilon^\chi_1, \;\text{col}^\chi_0 \rangle_{\text{Nek}} = \widetilde {\mathcal L}'_p (1, \chi),\) where \(\varepsilon_1\) is the (tame) elliptic unit introduced by Bley op. cit. and \(\varepsilon_1^\chi\) its \(\chi\)-part; \(\text{col}_0^\chi \in \widetilde H^1_f (K, T^\ast)\) is obtained via Coleman's theory as in case 1); \(\widetilde {\mathcal L}_p (s, \chi) = { \rho^{1-s}_{\text{cyc}} (\gamma) - 1 \over {1 \over p}\log_p \;\rho_{\text{cyc}} (\gamma)} \ldotp {\mathcal L}_p (s, \chi)\) is the imprimitive (one-variable) Katz \(p\)-adic \(L\)-function, where \({\mathcal L}_p (s, \chi)\) is the restriction of the two-variable \(p\)-adic \(L\)-function to \(\Gamma = \text{Gal} (K_\infty/K),\) \(K_\infty\) denoting the unique \({\mathbb Z}_p\)-extension of \(K\) which is unramified outside \({\mathcal P}.\)