Yuri
Bilu
U Bordeaux
Diversity in Parametric Families of Number Fields
Abstract.
Let $X$ be a projective curve over $\mathbb{Q}$ of genus $g$ and $t$ a non-constant $\mathbb{Q}$-rational function of degree $m>1$. For every $n\in \mathbb{N}$, pick $P_n\in X$ with $t(P_n)=n$. Hilbert's Irreducibility Theorem (HIT) says that for infinitely many $n$ the field $\mathbb{Q}(P_n)$ is of degree $m$ over $\mathbb{Q}$. Moreover, this holds for overwhelmingly many $n$: Among the number fields $\mathbb{Q}(P_1), \dots ,\mathbb{Q}(P_n)$ there is only $o(n)$ fields of degree less than $m$.
However, HIT does not say how many distinct field there are among $\mathbb{Q}(P_1), ... ,\mathbb{Q}(P_n)$. A 1994 result of Dvornicich and Zannier implies that for large $n$, there are at least $cn/\log n$ distinct among those fields, with $c=c(m,g)>0$. Conjecturally there should be a positive proportion (that is, $cn$) of distinct fields. This conjecture is proved in many special cases in the work of Zannier and his collaborators, but in general, getting rid of the log term seems very hard.
We make a little step towards proving this conjecture. While we cannot remove the log term altogether, we can replace it by log n raised to a power strictly smaller than 1. To be precise, we prove that for large $n$ there are at least $n/(\log n)^{1-e}$ distinct fields, where $e=e(m,g)>0$. A joint work with Florian Luca.