On Weyl sums for smaller exponents (Q2428767): Difference between revisions

Let \(f_k(\alpha;P)=\sum_{1\leq x\leq P}e(\alpha x^k)\) and \(g(\alpha;P)=\sum_{1\leq x\leq P}e(\alpha_1 x+\alpha_2 x^2+\cdots+\alpha_k x^k)\), where \(e(x)=e^{2\pi i x}\). Estimates for the mean value \(J_{s,k}(P)\) defined by \[ J_{s,k}(P)=\int_{[0,1)^k}|g(\alpha;P)|\,d\alpha, \] take the form \(J_{s,k}(P)\ll P^{2s-k(k+1)/2+\Delta_{s,k}}\), where \(\Delta_{s,k}\) is a real number depending on, at most, the positive integers \(s\) and \(k\). We say that the exponent \(\Delta_{s,k}\) is permissible if the above estimate holds for all real numbers \(P\). Let \(\mathfrak{m}_\theta\) denote the set of real numbers \(\alpha\) having the property that, whenever \(a\in\mathbb{Z}\) and \(q\in\mathbb{N}\) satisfying \(\gcd(a,q)=1\) and \(|q\alpha-a|\leq P^{\theta-k}\), then one has \(q>P^\theta\). In the paper under review, the authors show that if \(k\geq 4\) is a natural number and the exponent \(\Delta_{s,k-1}\) is permissible for \(s\geq k\), then for each \(\varepsilon>0\) one has \[ \sup_{\alpha\in\mathfrak{m}_1}|f_k(\alpha;P)|\ll P^{1-\sigma_k+\varepsilon}, \] where \(\sigma_k=\max_{s\geq k}(\frac{3-\Delta_{s,k-1}}{6s+2})\). More precisely, when \(10\leq k\leq 20\) this estimate holds with \(\sigma_k=1/\rho(k)\), where \(\rho(10)=440.87\), \(\rho(11)=575.81\), \(\rho(12)=733.58\), \(\rho(13)=910.41\), \(\rho(14)=1111.15\), \(\rho(15)=1331.61\), \(\rho(16)=1576.42\), \(\rho(17)=1841.79\), \(\rho(18)=2132.47\), \(\rho(19)=2444.02\), and \(\rho(20)=2781.54\). As an application related to the so-called Waring's problem, they show validity of the explicit upper bound \(\widetilde{G}(k)\leq H(k)\) with \(H(9)=365\), \(H(10)=497\), \(H(11)=627\), \(H(12)=771\), \(H(13)=934\), \(H(14)=1112\), \(H(15)=1307\), \(H(16)=1517\), \(H(17)=1747\), \(H(18)=1992\), \(H(19)=2255\), \(H(20)=2534\), where the function \(\widetilde{G}(k)\) denote the least integer \(t\) with the property that, for all \(s\geq t\), and all sufficiently large natural numbers \(n\), one has the asymptotic formula \[ R_{s,k}(n)=\frac{\Gamma(1+1/k)^s}{\Gamma(s/k)}\mathfrak{S}_{s,k}(n)n^{s/k-1}+o(n^{s/k-1}), \] where \(R_{s,k}(n)\) denote the number of representations of the natural number \(n\) as the sum of \(s\) \(k\)th powers of positive integers, for natural numbers \(s\) and \(k\), and \[ \mathfrak{S}_{s,k}(n)=\sum_{q=1}^\infty\sum_{\substack{ a=1\\ (a,q)=1}}^q \left(q^{-1}\sum_{r=1}^q e(ar^k/q)\right)^s e(-na/q). \]