Prime polynomial values of linear functions in short intervals (Q2257327)

The paper is mainly related with the prime number theorem and the Hardy-Littlewood prime tuple conjecture. For the first, it is conjectured that if \(I\) is an interval of length \(x^{\varepsilon}\), \(\varepsilon >0\), for \(x\) large enough, the number of primes is \(\sim \int_I\frac{dt} {\log t}\sim \frac{x^{\varepsilon}}{\log x}\). The analogue for function fields is that \(\sum_{f\in I} {\mathbf 1}(f)=\frac{\# I}{k}(1+O_k(q^{-1/2}))\) where \[ {\mathbf 1}(f)=\left\{ \begin{matrix} 1 &\text{if \(f\) is monic irreducible}\cr 0 & \text{otherwise}\end{matrix}\right. \] and we let \(\|f\|=q^{\deg f}\), \(I=I(f_0,\varepsilon)=\{f\in{\mathbb F}_q[t] : \|f-f_0\|\leq \|f_0\|^{\varepsilon}\}\), and \(f_0\) is a monic polynomial of degree \(k\), \(\frac{3}{k}\leq \varepsilon <1\). \textit{L. Rosenzweig} and the authors proved this result [Duke Math. J. 164, No. 2, 277--295 (2015; Zbl 1395.11132)]. The Hardy-Littlewood prime tuple conjecture asserts that \[ {\mathbf 1}(h+a_1)\cdots {\mathbf 1}(h+a_n)\sim {\mathcal G}(a_1,\ldots,a_n) \frac{x}{(\log x)^n},\quad x\to \infty, \] where \({\mathcal G}(a_1,\ldots,a_n)= \prod_p \frac{1-\nu(p)p^{-1}}{(1-p^{-1})^n}\) and \(\nu(p) =\# \{h\bmod p: (h+a_1)\cdots(h+a_n)\equiv 0 \bmod p\}\). For the function field case, the second author showed in [Int. Math. Res. Not. 2014, No. 2, 568--575 (2014; Zbl 1296.11165)] that for any \(k>0\) \[ \sum_{f\in{\mathbb F}_q[t] \text{\;monic}\atop \deg f=k} {\mathbf 1}(f+a_1)\cdots {\mathbf 1}(f+a_n)=\frac{q^k}{k^n} (1+O_{k,n}(q^{-1/2})) \] uniformly on all \(a_1,\ldots, a_n\in {\mathbb F}_q[t]\) with \(\deg(a_i)<k\) and \(q\) odd. The main result in this work is the following: Let \(B>0\) and \(0<\varepsilon<1\). Then the asymptotic formula \[ \sum_{f\in I(f_0,\varepsilon)} {\mathbf 1}(L_1(f))\cdots {\mathbf 1}(L_n(f)) = \frac{\# I(f_0,\varepsilon)}{\prod_{i=1}^n \deg (L_i(f_0))} (1+O_B(q^{-1/2})) \] holds uniformly for \(q\) odd, \(1\leq n\leq B\), distinct primitive linear functions \(L_1(X),\ldots,L_n(X)\) defined over \({\mathbb F}_q(t)\) each of height at most \(B\), and \(f_0\) a monic polynomial such that \(\frac{2}{\varepsilon}\leq \deg f_0\leq B\). Here \(L_i(X)=f_i(t)+g_i(t)X\) with \(f_i,g_i\in{\mathbb F}_q[t]\), \(g_i\neq 0\) and \(\mathrm{height}(L_i):=\max\{\deg f_i,\deg g_i\}\). The proof is based in the computation of the Galois groups of some polynomials. This is carried out in Section 2. In Section 3 a theorem more general than the main result is proved. The proof is in the spirit of several similar results once one has the Galois group. In the last section it is presented a discussion on primes in short intervals and also on arithmetic progressions are considered.