TITLE: Structure of $E(Q_p)$ for elliptic curves with anomalous reduction modulo $p$
QUESTION [15 upvotes]: For simplicity, take $p\ge7$ a prime and $E/\mathbb{Q}$ an elliptic curve with good anomalous reduction at $p$, i.e., $|E(\mathbb{F}_p)|=p$. There is a standard exact sequence for the group of points over $\mathbb{Q}_p$,
$$
  0 \to \hat{E}(p\mathbb{Z}_p) \to E(\mathbb{Q}_p) \to E(\mathbb{F}_p) \to 0.
$$
The assumption that $p$ is anomalous implies that $E(\mathbb{F}_p)=\mathbb{Z}/p\mathbb{Z}$, while the formal group $\hat{E}(p\mathbb{Z}_p)$ is isomorphic to the formal group of the additive group, so we obtain an exact sequence
$$
  0 \to p\mathbb{Z}_p^+ \to E(\mathbb{Q}_p) \to \mathbb{Z}/p\mathbb{Z} \to 0.
  \qquad(*)
$$
In work I'm doing on elliptic pseudoprimes, the question of whether the sequence $(*)$ splits has become relevant. Questions:

Are there places in the literature
where the split-versus-nonsplit
dichotomy for anomalous primes comes
up, e.g., in the theory of $p$-adic
modular forms? 
Is there a name for
this dichotomy in the literature?
Aside from the obvious observation
that the sequence $(*)$ splits if and only
if $E(\mathbb{Q}_p)$ has a
$p$-torsion point, are there other
natural necessary or sufficient
conditions for splitting?

REPLY [4 votes]: If you work modulo $p^2$, the sequence $0 \to p\mathbb{Z}/p^2 \to E(\mathbb{Z}/p^2) \to E(\mathbb{Z}/p) \to 0$, splits iff $E$ is a canonical lift (as in Álvaro's answer). This comes up in cryptography (!) in the Smart-Satoh-Araki attack on the ECDLP for anomalous curves. This point is discussed in the Satoh-Araki paper, I believe.