Twisted triple product root numbers and a cycle of Darmon-Rotger
David T. -B. G. Lilienfeldt
TL;DR
The paper studies the Darmon–Rotger cycle on the triple product X_0(p)^3 over the quadratic field K ramified at p, proving the cycle Xi = Δ_+ − Δ_- is null-homologous and analyzing its symmetry. By constructing and analyzing twisted triple product L-functions L(F,χ,s) with χ the Legendre symbol at p, it shows the global root number W(F,χ) = −1, implying L(F,χ,s) has odd-order vanishing at the center s = 2. Under Beilinson–Bloch–Kato conjectures, this supports the expectation that the Darmon–Rotger cycle could be non-torsion, aligning with the Gross–Zagier philosophy. The work also develops a detailed, representation-theoretic framework via Weil–Deligne representations and local epsilon factors to establish these root-number phenomena, and outlines natural generalizations to composite levels and higher weights alongside connections to p-adic families and generalized Gross–Kudla–Schoen cycles.
Abstract
We consider an algebraic cycle on the triple product of the prime level modular curve $X_0(p)$ with origins in work of Darmon and Rotger. It is defined over the quadratic extension of $\mathbb{Q}$ ramified only at $p$ whose associated quadratic character $χ$ is the Legendre symbol at $p$. We prove that it is null-homologous and describe actions of various groups on it. For any three normalised cuspidal eigenforms $f_1, f_2, f_3$ of weight $2$ and level $Γ_0(p)$, we prove that the global root number of the twisted triple product $L$-function $L(f_1\otimes f_2\otimes f_3\otimes χ, s)$ is $-1$. Assuming conjectures of Beilinson and Bloch, and guided by the Gross-Zagier philosophy, this suggests that the Darmon-Rotger cycle could be non-torsion, although we do not currently have a proof of this.
