Fix $p>0$ a rational prime, and $K$ an algebraic number field with Galois group $G_K:=Gal(\bar{\mathbb{Q}}/K) $. The Fontaine-Mazur conjecture predicts that if $\rho:G_K\rightarrow GL(V)$ is a finite dimensional $\mathbb{Q}_p$-representation, then it comes from a motive over $K$ (like a subquotient of $H^i_c(X\times_K\bar{\mathbb{Q}}, \mathbb{Q}_p)$) exactly when it is unramified over almost every finite place, and potentially semi-stable over those finite prime dividing $p$.
My question is the contrary: how many examples do we have for $p$-adic Galois representations having infinite images but for which the conditions of Fontaine-Mazur fails? Maybe it should be difficult to construct them when the dimension of the representation is large? Could one get continuous $p$-adic representations that ramifies at infinitely many places?
Best Answer
Here are two things that can occur:
1) If V is the representation attached to an overconvergent modular form f, then V will be unramified at almost every prime but will not be de Rham at p (unless f is classical).
2) Ramakrishna has written an article "Infinitely ramified Galois representations". Here's part of the introduction: "In this paper we show how to construct [...] representations [...] that are ramified at an infinite number of primes." Under GRH, these repns are crystalline at p.
So both conditions in FM can fail independently.