On the Birch and Swinnerton-Dyer formula modulo squares for certain quadratic twists of elliptic curves

Abstract: Let $E/\mathbb{Q}$ be an elliptic curve with conductor $N=N_+N_-$, where $N_+$ and $N_-$ are coprime and $N_-$ is squarefree. Let $D$ be a positive fundamental discriminant satisfying the modified Heegner hypothesis with respect to $(N_+,N_-)$: primes dividing $N_+$ (resp. $N_-$) split (resp. are inert) in $\mathbb{Q}(\sqrt{D})$; we denote by $E^D/\mathbb{Q}$ the quadratic twist of $E/\mathbb{Q}$ by $D$. In the first half of the paper we consider the situation where $N_-$ is a squarefree product of an odd number of distinct primes, and we show the following: assuming that $E/\mathbb{Q}$ is of analytic rank zero (resp. one), and that the Birch and Swinnerton-Dyer formula holds for $E/\mathbb{Q}$ modulo $(\mathbb{Q}^{\times})^2$, then for those $D$ such that $E^D/\mathbb{Q}$ is of analytic rank one (resp. zero), we also have the validity of the Birch and Swinnerton-Dyer formula for $E^D/\mathbb{Q}$ modulo $(\mathbb{Q}^{\times})^2$. To show this, we establish auxiliary results without rank assumptions. The most difficult case is when $D$ is even, and our proof crucially relies on the recent classification of how local Tamagawa numbers change under quadratic twists. In the final part of the paper analogous results are also obtained in the other situation when $N_-$ is a squarefree product of an even number distinct primes, concerning the case when both $E/\mathbb{Q}$ and $E^D/\mathbb{Q}$ have analytic rank zero (resp. one). As a consequence of our work, we obtain that if $E/\mathbb{Q}$ is semistable with conductor $N$ and whose analytic rank is at most one, then for any positive fundamental discriminant $D$ that is coprime to $N$, such that $E^D/\mathbb{Q}$ again has analytic rank at most one, we have that the Birch and Swinnerton-Dyer formula modulo $(\mathbb{Q}^{\times})^2$ holds for $E/\mathbb{Q}$ if and only if it holds for $E^D/\mathbb{Q}$.