An infinite family of elliptic curves and Galois module structure (Q1306232)

This pretty paper shows that a conjecture of Martin Taylor, concerning a Galois module problem attached to elliptic curves, and proved by Agboola and Pappas independently for \(p>3\), admits an infinite family of counterexamples for \(p=2\). Taylor's conjecture says: For \(E\) an elliptic curve with good reduction everywhere over a number field \(F\), for \(p\) a prime and \(i\geq 1\), a certain ``Kummer order'' obtained (roughly speaking) by dividing any point \(P\in E_{\text{tors}}\) by \([p^i]\) is free over the relevant Hopf algebra \({\mathfrak A}_i\), the Cartier dual of the Hopf algebra representing the \([p^i]\)-torsion subscheme of \(E\). Note that this construction already uses that the curve has good reduction everywhere. In order to establish the counterexamples, the authors use a family of elliptic curves which was found by Setzer in 1978. (In the present paper, some oversights in Setzer's formulas are corrected.) The curves are defined over quadratic imaginary fields \(k\) of the form \({\mathbb Q} (\sqrt{-65m_1})\) with \(m_1\) a square mod 5 and mod 13, and 65 a square mod \(m_1\). The authors calculate the Kummer order and the algebra \({\mathfrak A}_i\) in this setup, \(P\) being the nontrivial \(k\)-rational 2-division point \((0,0)\) on \(E\), with \(p=2\) (as we said) and \(i=1\). Among other things, one needs the etale-connected decomposition of the group scheme \(E[2]\) at primes over 2. In the end, the nonfreeness of the Kummer order over \({\mathfrak A}_i\) comes down to the criterion that the ideal \({\mathfrak p}_5{\mathfrak p}_{13}\) should not be principal in \(k\). This is the case iff \(m_1\) is not 1 (just look at the equation \(4\cdot 65=a^2+65m_1b^2\)), so one does get an infinite family of counterexamples.