The derived Hecke algebra for dihedral weight one forms (Q2083497)

A conjecture proposed by Venkatesh and others posits that when spaces of automorphic forms show up in different degrees of cohomology, they are related by the action of a motivic cohomology group. The paper [\textit{M. Harris} and \textit{A. Venkatesh}, Exp. Math. 28, No. 3, 342--361 (2019; Zbl 1480.11049)] formulates an explicit version of this conjecture in the case of weight \(1\) modular forms. The present paper then proves the Harris-Venkatesh conjecture in the \textit{dihedral} case under additional ramification assumptions. We now describe in slightly more detail the Harris-Venkatesh conjecture and the contents of this paper. The Harris-Venkatesh conjecture computes a pairing \(\langle G, \mathscr{G} \rangle\) arising from Serre duality. The element \(G \in H^0(X^0(N), \Omega^1)\) is a weight \(2\) level \(N\) cusp form given as the trace \(X_0(Nd) \to X_0(N)\) of \(g(z)g^*(NZ) \in S_2(Nd)\), where \(g \in H^0(X_1(d), \omega)\) is a weight \(1\) level \(d\) newform and \(g^*\) is the newform whose Fourier expansion is the complex conjugate of that of \(g\). The element \(\mathscr{G} \in H^1(\overline{X}, \mathcal{O}_{\overline{X}})\) (where \(\overline{X} = X_0(N)_{\mathbb{Z}/p^t\mathbb{Z}}\) and \(p^t\) is the highest power of \(p\) dividing \(N-1\)) is the \emph{Shimura Class}, coming from the class in \(H^1_{\mathrm{et}}(X_0(N), (\mathbb{Z} / N \mathbb{Z})^{\times} \otimes \mathbb{Z}[\frac{1}{6}])\) corresponding to the cover \(X_1(N) \to X_0(N)\) via a fixed discrete logarithm \(\mathrm{log}: (\mathbb{Z}/N\mathbb{Z})^{\times} \to \mathbb{Z} / p^t \mathbb{Z}\). Then the Harris-Venkatesh conjecture says this pairing is equal (up to constant independent of \(N\) and \(p\)) to \(\mathrm{log}(\mathrm{red}_N(u_g))\), the discrete logarithm of the modulo \(N\) reduction of a certain unit \(u_g\) attached to \(g\). To carry out the computation, they transport the computation to a suitable quaternion algebra \(B\). In particular, there is a theta lift \(\Theta: \text{modular forms on } B \to S_2(N)\) and (up to something pairing trivially with \(\mathscr{G}\)) one has \(G = \Theta(Z_{K,\psi})\) where \(Z_{K,\psi}\) is a certain Heegner cycle, equal to a formal combination of supersingular points in characteristic \(N\). Dually, one has \(\Theta^*(\mathscr{G}) = \mathfrak{U}_N\) where \(\mathfrak{U}_N\) is an explicit \textit{higher Eisenstein element} as in [\textit{E. Lecouturier}, Invent. Math. 223, No. 2, 485--595 (2021; Zbl 1472.11134)]. Hence, one reduces to evaluating \(\langle Z_{K, \psi} , \mathfrak{U}_N \rangle\) which is possible due to explicit constructions of units available in the dihedral case.