An Erdős-Kac law for local solubility in families of varieties (Q2036432)

Let \(V\) be a smooth projective variety over \(\mathbb{Q}\), and \(\pi :\, V\to\mathbb{P}^n\) a dominant morphism over \(\mathbb{Q}\), with geometrically integral generic fibre. The authors are interested in the behaviour of the number \(\omega_{\pi}(x):=\#\{ p\ \text{prime}:\,\pi^{-1}(x)(\mathbb{Q}_p )=\emptyset\}\), as \(x\) varies through \(\mathbb{P}^n(\mathbb{Q})\). This number is known to be finite for all \(x\) outside a proper Zariski-closed subset of \(\mathbb{P}^n\). \textit{D. Loughran} and \textit{A. Smeets} [Geom. Funct. Anal. 26, No. 5, 1449--1482 (2016; Zbl 1357.14028)] introduced a quantity \(\Delta (\pi )>0\) and proved among other things that \(\Delta (\pi )=0\) precisely if \(\omega_{\pi}(x)=0\) for a positive proportion of \(x\). The authors make this more precise by proving an Erdős-Kac type theorem for \(\omega_{\pi}\) in the case that \(\Delta (\pi )>0\), and a limit law if \(\Delta (\pi )=0\). To formulate their theorems, we introduce the necessary notation: -- for \(x\in \mathbb{P}^n(\mathbb{Q})\) let \(\omega_{\pi }(x)\) denote the number of primes \(p\) such that \(\pi^{-1}(x)(\mathbb{Q}_p )=\emptyset\); this quantity is finite for all \(x\) outside a proper Zariski-closed subset; -- for a scheme-theoretic point \(x\) of \(\mathbb{P}^n\), denote by \(\kappa (x)\) its residue field, \(\overline{\kappa (x)}\) an algebraic closure and \(\pi^{-1}(x)_{\overline{\kappa (x)}}:=\pi^{-1}(x)\otimes_{\kappa (x)}\overline{\kappa (x)}\). Suppose that the action of \(\mathrm{Gal}(\overline{\kappa (x)}/\kappa (x))\) on the collection of irreducible components of \(\pi^{-1}(x)_{\overline{\kappa (x)}}\) of multiplicity \(1\) factors through a finite group \(\Gamma_x\). Then define \[ \delta_{x}(\pi ):=\frac{\{\#\text{\(\gamma\in\Gamma_x\): \(\gamma\) fixes an irreducible component of \(\pi^{-1}(x)_{\overline{\kappa (x)}}\) of multiplicity \(1\)}\}}{\#\Gamma_x} \] and subsequently \(\Delta (\pi ):=\sum_{x\in X^{(1)}}(1-\delta_x(\pi ))\), where \(X^{(1)}\) is the set of codimension \(1\) points of \(\mathbb{P}^n\); -- define the height of \(x\in\mathbb{P}^n(\mathbb{Q})\) by \(H(x):=\max (|x_0|,\ldots ,|x_n|)\) where \(x_0,\ldots , x_n\in\mathbb{Z}\) are coordinates of \(x\) with \(\gcd (x_0,\ldots ,x_n)=1\); -- lastly, for \(B\in\mathbb{R}_{\geq 1}\) and a subset \(A\) of \(\mathbb{P}^n(\mathbb{Q})\) define \(\nu_B(A):=\#\{ x\in A:\, H(x)\leq B\}/\#\{ x\in\mathbb{P}^n(\mathbb{Q}):\, H(x)\leq B\}\). Theorem 1. Assume that \(\Delta (\pi )>0\). Then for any \(a,b\in\mathbb{R}\) with \(a\leq b\) we have \[ \lim_{B\to\infty}\nu_B\left(\left\{ x\in\mathbb{P}^n(\mathbb{Q}): a\leq \frac{\omega_{\pi}(x)-\Delta (\pi )\log\log H(x)}{\sqrt{\Delta (\pi )\log\log H(x)}}\leq b\right\}\right)=\frac{1}{\sqrt{2\pi}}\int_a^b e^{-t^2/2}dt. \] The authors prove this by more or less showing that \(\frac{\omega_{\pi}(x)-\Delta (\pi )\log\log H(x)}{\sqrt{\Delta (\pi )\log\log H(x)}}\) has the same moments as the standard normal distribution, using a result of \textit{A. Granville} and \textit{K. Soundararajan} [NATO Sci. Ser. II, Math. Phys. Chem. 237, 15--27 (2007; Zbl 1145.11071)]. The authors also consider the case where \(\Delta (\pi )=0\) (the so-called 'pseudo-split case'). Here, they study the quantities \[ \tau_{\pi} (j,B):=\frac{\{ x\in \mathbb{P}^n(\mathbb{Q}):\, H(x)\leq B,\, \pi^{-1}(x)\ \text{smooth,}\ \omega_{\pi}(x)=j\}}{\#\{ x\in\mathbb{P}^n(\mathbb{Q}):\, H(x)\leq B\}}. \] Theorem 2. The function \(\tau_{\pi}:\,\mathbb{Z}_{\geq 0}\to \mathbb{R}:\, \tau_{\pi}(j):=\lim_{B\to\infty} \tau_{\pi} (B,j)\) is well-defined and defines a probability measure on \(\mathbb{Z}_{\geq 0}\) and moreover, \[ \tau_{\pi}(j)\ll_{\pi}\frac{1}{(1+j)^j(\log (2+j))^{j/2}}\ \ \text{for }j\in\mathbb{Z}_{\geq 0} \] where the implied constant depends only on \(\pi\). The authors' proof builds further on arguments of \textit{M. J. Bright} et al. [Compos. Math. 152, No. 7, 1435--1475 (2016; Zbl 1348.14067)]. The authors illustrate their results by a couple of examples.