3
$\begingroup$

It is known that the complex isomorphism class of an elliptic curve $E/\mathbb{C}$ is uniquely determined by its $j$-invariant. One way to define it algebraically for a curve

$$\displaystyle E : y^2 = x^3 + Ax + B, A,B \in \mathbb{C}$$

is by

$$j(E) = 1728 \cdot \frac{4A^3}{4A^3 + 27B^2}.$$

Of course $j$ is also a modular function, and can be expressed as $j(\tau)$, where $\tau$ is a complex number such that $E/\mathbb{C} \cong \mathbb{C}/\langle 1, \tau \rangle$. By reduction theory, we can further assume that $\tau$ is in the Gauss fundamental domain in the upper half-plane: that is, we may assume that $|\Re(\tau)| \leq 1/2$ and $\Im(\tau) \geq 1$.

If we assume that $A,B \in \mathbb{Z}$ (or more generally, simply algebraic integers), then we can attach a modular height to the elliptic curve by simply evaluating the Weil height (as an algebraic number) of the $j$-invariant of $E$, say $h(j(E))$.

When $\gcd(A,B) = 1$, then $h(j(E))$ is typically close to the so-called naive height of $E$, namely $H(E_{A,B}) = \max\{4|A|^3, 27B^2\}$. However, there are cases when the two heights differ significantly.

Another way to measure the size of $j(E)$ with $E$ defined over $\mathbb{Q}$ is to simply take its absolute value. In this case, we see that if we expect $4A^3$ and $4A^3 + 27B^2$ to be roughly the same size, then $|j(E)|$ is typically very small.

Therefore, elliptic curves over $\mathbb{Q}$ whose $j$-invariants have very large absolute value are exceptional, as they correspond to those curves whose discriminants are very small compared to the naive height of the curve.

My question is, what arithmetic information can one glean from a curve $E/\mathbb{Q}$ with $\gcd(A,B) = 1$ (or more generally, taking a minimal element in the twist family) whose $j$-invariant has very large absolute value? Alternatively, such an elliptic curve has a period $\tau$ in the fundamental domain with large imaginary part.

$\endgroup$
8
  • 1
    $\begingroup$ Why not choosing $y^2= 4x^3-27j/(j-1728)x-27j/(j-1728)$, the arithmetic complexity is about how it differs from the integer equation with minimal discriminant $\endgroup$
    – reuns
    Commented Jul 19, 2019 at 5:07
  • $\begingroup$ It is more cautious to talk about what arithmetic information one can gain for the family of (quadratic) twists of elliptic curves $E/\mathbb{Q}$ with that $j$-invariant. Or do you choose one curve in that family? $\endgroup$ Commented Jul 19, 2019 at 8:21
  • 2
    $\begingroup$ In the $p$-adic version, large $\vert j(E)\vert _p$ measures if $E$ has potential multiplicative reduction and how many components it has. $\endgroup$ Commented Jul 19, 2019 at 8:23
  • 1
    $\begingroup$ Roughly speaking, large $|j|_p$ means that $E$ has "highly" multiplicative reduction at $p$ (and thus it's Neron polygon has lots of sides). So large $|j|_\infty$ means that $E$ has "highly multiplicative reduction at the infinite place of $\mathbb Q$." One can feed that into various machines to get information. For example, Lang's height lower bound conjecture $\hat h(P)\gg\log|\Delta|$ is true, I think. Is that the sort of thing that you had in mind? $\endgroup$ Commented Jul 19, 2019 at 15:16
  • 1
    $\begingroup$ Which phenomena? The $p$-adic fact is standard, the number of components on the Neron model is the valuation of the discriminant. The height result probably isn't written down anywhere, and I'm not 100% sure it's correct; but if it is, it should follow from the methods in my 1981 Duke Math J paper, since one can pigeon-hole to get $\hat h(P)$ larger than a multiple of $\log|j|$, and if $|j|$ is big enough, that will overwhelm the worst possible contribution of $-\frac1{24}\log\Delta$ coming from the finite places. If you have an application ,I could work it out more carefully. $\endgroup$ Commented Jul 19, 2019 at 18:10

0

You must log in to answer this question.

Browse other questions tagged .