For an elliptic curve $E/\mathbb{Q}$, let us denote by $\Delta_{\min}(E)$ the minimal discriminant of $E$ and $N(E)$ the conductor of $E$. Then it is well-known that $N(E) | \Delta_\min(E)$.

The Szpiro ratio of $E$ is defined as the ratio $$\displaystyle \beta_E = \frac{\log |\Delta_\min(E)|}{\log N(E)}.$$

This definition is motivated by Szpiro's conjecture, which asserts that for all $\epsilon > 0$ there exists only finitely many curves $E/\mathbb{Q}$ satisfying

$$\displaystyle \beta_E > 6 + \epsilon.$$

Of course, it is known that Szpiro's conjecture is equivalent to the $abc$-conjecture.

Note that the value of $6$ in the conjecture is sharp, because one can construct explicit families with Szpiro ratio converging to $6$. For example, if $E$ is given by a short Weierstrass model

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

with the property that for all primes $p$ we have $p^4 | A \Rightarrow p^6 \nmid B$, then the discriminant is given by $\Delta(E) = 16(4A^3 - 27B^2)$. If $\Delta(E)$ is $12$-power-free, then $\Delta(E) = \Delta_\min(E)$. We can then find an integer $c_1$ such that the equation

$$\displaystyle 16(4x^3 - 27y^2) = c_1 z^6$$

defines an elliptic generalized Fermat equation with a solution, which then gives at least one family of parametrized solutions. In this family, the Szpiro ratio will approach $6$.

Similar constructions can be given for the following families: elliptic curves with a rational 2-torsion point, with 3 rational 2-torsion points, and those with an abelian 2-torsion field.

My question is: is the Szpiro ratio $\beta_E$, as $E$ varies over all $E/\mathbb{Q}$, dense in the interval $[1,6]$?

  • 4
    $\begingroup$ Well, you can start with a curve with large conductor $N$ and Szpiro ratio $>6$ and then quadratic twist by suitably chosen primes (coprime to $N$) $p$ so that $p^2$ shows up in the conductor and $p^6$ in the minimal discriminant. With care, this should give density in $[3,6)$. Starting with large conductor and Szpiro ratio $1$, then the same argument gives density in $(1,3)$. $\endgroup$ Feb 20, 2022 at 0:38
  • $\begingroup$ @MikeBennett I think that does indeed answer my question, if you'd like please write it as an answer and I will accept. Thanks! $\endgroup$ Feb 23, 2022 at 17:56

1 Answer 1


The answer is "yes". Use one of the various families with Szpiro ratio exceeding $6$ and large conductor $N$, and then twist by a prime $p$ of appropriate size, coprime to $N$ (which increases the conductor by $p^2$ and the minimal discriminant by $p^6$). With a certain amount of care, this gives density in the interval $[3,6]$. A similar argument, starting with a curve of Szpiro ratio arbitrarily close to $1$ (there are various ways to construct these) and large conductor then gives density in $[1,3]$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.