# What does the Tate module of an elliptic curve tell us?

I started studying elliptic curves, and I see that it is rather common to take the Tate module of an elliptic curve (or, of the Jacobian of a higher genus curve).

I'm having a hard time isolating the benefits of this construction. Can you give an example (or claim) explaining how the Tate module sheds light on some properties of the curve?

• You should begin by carefully understanding, in the case that $E = \mathbb C/\Lambda$ is an elliptic curve over the complex numbers, the canonical isomorphism between the $\ell$-adic Tate module and $\mathbb Z_{\ell} \otimes_{\mathbb Z} \Lambda$ (or, what is the same, t – tracing Nov 30 '14 at 6:56
• In an abelian variety, it's part of the homotopy group. – user40276 Nov 30 '14 at 7:12

You should begin by carefully understanding, in the case that $E = \mathbb C/\Lambda$ is an elliptic curve over the complex numbers, the canonical isomorphism between the $\ell$-adic Tate module and $\mathbb Z_{\ell} \otimes_{\mathbb Z} \Lambda$ (or, what is the same, the inverse limit $\varprojlim_n \Lambda/\ell^n\Lambda$).

Once you've understood that, you could read the proofs in Silverman classifying the possible endomorphism rings for elliptic curves. Silverman's proofs work in arbitrary characteristic, and use Tate modules. But you could try to imitate them for elliptic curves over $\mathbb C$, using the lattice $\Lambda$ directly as a tool. The arguments then become quite a bit simpler. Comparing these simple arguments with the Tate module arguments should further build up your intuition.

The next step is to learn the proof of the Hasse--Weil theorem counting points on elliptic curves over finite fields, and the philosophy of thinking of its an application of the Lefschetez fixed point theorem for the Frobenius endomorphism --- with the Tate module playing the role of $H_1$. If you recall that in the complex case the lattice $\Lambda$ is canonically identified with $H_1(E,\mathbb Z)$, this will add yet more intuition.

Summary: Tate modules substitute for the lattice $\Lambda$ which plays such an important role in the study of elliptic curves over $\mathbb C$.

• +1 This is also a good answer! I totally forgot to mention this great piece of intuition, which I was also told once by a wise mathematician. – Alex Youcis Nov 30 '14 at 7:53

There are many, many ways in which the knowledge of the various Tate modules gives you information about the elliptic curve. For example:

Theorem(Neron-Ogg-Shafarevich): Let $E$ be an elliptic curve over a number field $K$. Let $p$ be a prime of $\mathbb{Q}$. Then, $E$ has good reduction at $\mathfrak{p}$, for $p\nmid\mathfrak{p}$, if and only if the representation $\rho_p:G_K\to \text{GL}_2(\mathbb{Z}_p)$ is unramified at $\mathfrak{p}$.

Here $G_K$ is the absolute Galois group of $K$, and $\rho_\mathfrak{p}$ is any of the representations of the form

$$G_{K_\mathfrak{p}}\to G_K\to \text{GL}_2(\mathbb{Z}_p)$$

where $G_K\to \text{GL}_2(\mathbb{Z}_p)\cong \text{GL}_2(T_p E)$ is the representation coming from the action of $G_K$ on $T_p E$. Also, being unramified means that this representation's kernel contains the inertia group $I_\mathfrak{p}$.

In fact, you can recover almost all the local $L$-factors of an elliptic curve over a number field from its Tate modules (and its Galois actions), which, by a hard theorem of Tate/Faltings, actually determines $E$ up to $\mathbb{Q}$-isogeny. So, in a very real sense, the Tate modules (and their associated Galois representations) determine everything about the elliptic curve, at least over a number field.

Let's start over the complex numbers, so that $$E$$ is an elliptic curve over $$\mathbb C$$. Then the natural map $$\mathrm{End}(E) \to \mathrm{End}(H^1(X,\mathbb C))$$ is injective. Suppose now that you have an elliptic curve over $$\mathbb F_p$$. What could the analogue of this statement be?

If $$E$$ is an elliptic curve over an arbitrary field $$k$$, and $$\ell$$ is a prime number invertible in $$k$$, then the Tate module $$\mathrm{V}_\ell(E) = T_{\ell} E \otimes \mathbb Q$$ is a replacement for $$H^1(X,\mathbb C)$$ (in some sense). For instance, for all prime numbers $$\ell$$ invertible in $$k$$, the natural map $$\mathrm{End}(E) \to \mathrm{End}_{\mathbb Q_\ell}( V_\ell E)$$ is injective.

Roughly speaking, the Tate module is the (dual of the) $$\ell$$-adic realization of the "motive" of $$E$$. In fact, it is more than just a $$\mathbb Q_\ell$$-vector space (of dimension two) as it carries a natural action of the absolute Galois group $$G_k = Gal(\bar k/k)$$ of $$k$$.

If $$k$$ is finitely generated over its prime field, then the natural map

$$\mathrm{End}(E) \otimes \mathbb Q_\ell \to \mathrm{End}_{\mathbb Q_\ell[G_k]}( V_\ell E)$$ is an isomorphism. This is due to Tate, Zarhin and Faltings.

So in arithmetic situations (e.g., number fields or finite fields), the Tate module tells you essentially everything about the motive.

Sometimes properties of the motive imply similar properties of the underlying variety. The criterion of Neron-Ogg-Shafarevich (see Alex Youcis's answer) is an important example of this phenomenon. In fact, the criterion of Neron-Ogg-Shafarevich tells you that an elliptic curve $$E$$ over a discretely valued field $$K$$ has good reduction over $$O_K$$ if and only if there exists a prime number $$\ell$$ with $$\ell \in k^\ast$$ such that the $$\ell$$-adic Tate module (read "$$\ell$$-adic realization of the motive of $$E$$") has good reduction over $$O_K$$ (in the sense that it is unramified as a Galois representation).

Another application comes from the before-mentioned theorem of Tate, Zarhin and Faltings. If $$E_1$$ and $$E_2$$ are elliptic curves over a finitely generated field $$k$$ and $$\ell \in k^\ast$$, then any $$G_k$$-equivariant morphism $$V_\ell E_1 \to V_\ell E_2$$ is induced by an element of $$\mathrm{Hom}(E_1,E_2)\otimes \mathbb Q_\ell$$. Therefore, the elliptic curves $$E_1$$ and $$E_2$$ are $$k$$-isogenous if and only if their $$\ell$$-adic Tate modules (read $$\ell$$-adic realizations of their motives") are isomorphic as Galois representations.

If $$k$$ is finitely generated of characteristic zero, you can also read of the system of Tate modules $$\{V_\ell(E)\}_{\ell}$$ whether your curve has complex multiplication.

Similar statements hold for abelian varieties.

Before you try to understand the above facts in full detail, I recommend you follow guy_in_seoul's advice and develop your intuition for the Tate module by studying elliptic curves over $$\mathbb C$$.

• Nice answer! Would you mind clarifying something for me? I get confused when people start talking about motives as being concrete objects. I thought for a scheme $X$ a good candidate for its associated motive is its Albanese scheme. If $X$ is a curve, then this is just the dual of the Jacobian of $X$. This meshes well with your answer over $\mathbb{C}$, since, at least, $\text{Alb}(E)$ is $H^1(E,\mathbb{C})$ But how precisely do I think about the Tate module as being the Albanese variety? If the Albanese variety is the wrong way of thinking about motives, please correct/enlighten me! Thanks! – Alex Youcis Nov 29 '14 at 21:19
• @AlexYoucis: The Albanese is just giving the motivic $H^1$ (or maybe $H_1$), not any higher cohomology. So only for objects like curves or abelian varieties (for which the $H^1$ generates the whole cohom. ring) is the Albanese capturing the entire motive. For many people (e.g. number theorists) a motive is most easily captured by its $\ell$-adic (co)homology. This is exactly what the Tate module is doing --- it is the $\ell$-adic $H_1$ of the elliptic curve (or more generally of an abelian variety). – tracing Nov 30 '14 at 6:55
• @guy-in-seoul Thanks for the response! Can I ask a silly question? What does $H_1$ mean? I know that $H^1(A,\mathbb{Q}_\ell)$ is $V_\ell(A)$, for an abelian variety $A$, but I have no idea what $H_1$ means. I would have assumed it was a typo, if not for the "motivic $H^1$ (or maybe $H_1$). Is there some notion of homology when one things motivically? Of all the specific cohomology theories I know (etale, de Rham/crystalline, etc.), none of them have homology theories, so I wouldn't have thought there is motivic homology, but perhaps I'm wrong. Thanks again! – Alex Youcis Nov 30 '14 at 7:55
• The etale $H^1$ of an abelian variety is not $V_{\ell}(A)$, but rather $V_{\ell}(A)^{\vee}$. (You can see this already in the complex case; the lattice $\Lambda$, which is covariantly related to the Tate module --- not contravariantly related --- is $H_1$ of the elliptic curve, not $H^1$). For any of the standard cohom. theories, you can define $H_i$ to be the dual of $H^i$. – tracing Nov 30 '14 at 13:43