Explicit relations between pair correlation of zeros and primes in short intervals (Q442501): Difference between revisions

Let \(\zeta(s)\) denote the Riemann zeta-function, and let \(\rho=\beta+i\gamma\) denote a nontrivial zero of \(\zeta(s)\). Montgomery's pair correlation function is defined as \[ F(X,T) = \sum_{0<\gamma, \gamma'\leq T} X^{i(\gamma-\gamma')} w(\gamma-\gamma'), \quad w(u)=\frac{4}{4+u^2} \] where the sum runs over two sets of ordinates of zeros of \(\zeta(s)\). The paper under review explores the equivalence between the asymptotic behavior of the function \(F(X,T)\) and the distribution of primes in short intervals extending and improving results of \textit{T. H. Chan} [J. Lond. Math. Soc. (2) 68 (2003) 579--598; Zbl 1054.11047]. To state the main results, write \[ F(X,T) = \frac{T}{2\pi} \left( \log \frac{T}{2\pi} - 1 \right) + R_F(X,T) \] and write \[ \int_X^{2X} \left( \psi(x+\theta x)-\psi(x)-\theta x \right)^2 dx = \frac{3}{2} X^2 \theta \left( \log(1/\theta)+C\right) + R_J(X,\theta), \] where \(\psi(x)=\sum_{n\leq x} \Lambda(n)\) is the usual summatory function of von Mangoldt's function \(\Lambda(n)\) and \(C=1-\gamma_0-\log 2\pi\) (\(\gamma_0\) is Euler's constant). Assuming the Riemann hypothesis (RH), the authors investigate the relationship between the error terms \(R_F(X,T)\) and \(R_J(X,\theta)\) with \(X, T,\) and \(\theta\) in suitable ranges. They prove: 1. Assume RH and let \(0<\alpha <1\), \(\beta\geq 0\), and \(1\leq A_1 \leq A_2\). If \[ R_F(X,T) \ll T^{1-\alpha} \log^{-\beta}T \] uniformly for \(X^{1/A_2}/\log^{(3+\beta)}X\leq T \leq X^{1/A_1}/\log^{(3+\beta)}X\), then \[ R_J(X,\theta) \ll X^2 \theta^{1+\alpha/2} \log^{(1-\beta)/2}(1/\theta) \] uniformly for \(X^{-1/A_1(1+\alpha)} \leq \theta \leq X^{-1/A_2(1-\alpha)}\). 2. Assume RH and let \(0<\alpha <1\), \(\beta\geq 0\), \((\alpha,\beta)\neq (0,0)\), and \(1\leq A_1 \leq A_2\). If \[ R_J(X,\theta) \ll X^2 \theta^{1+\alpha} \log^{-\beta}(1/\theta) \] uniformly for \(X^{-(\alpha/2+1)/A_1}/\log^{(3+\beta)}X \leq \theta \leq X^{-1/A_2}/\log^{4(\beta+2)/3}X\), then \[ R_F(X,T) \ll T^{1-\alpha/(\alpha+3)} (\log T)^{(\alpha-\beta+2)/(\alpha+3)} \] uniformly for \(X^{(\alpha+3)/3A_2(1-\alpha)} \leq T \leq X^{(\alpha+3)/3A_1}\). The authors also discuss the relationship between these results and the integral \[ \int_X^{2X} \left( \psi(x+h)-\psi(x)-h \right)^2 dx \] which they note is perhaps the ``more natural'' way to study primes in short intervals.