Discrepancy bounds for the distribution of the Riemann zeta-function and applications (Q2299596)

For any complex-valued function $f:{\mathbb{R}}\to {\mathbb{C}}$ and any rectangle $R\subset {\mathbb{C}}$, let $P_T$ denote the proportion function: \[ P_T(f(t)\in R) = \frac{1}{T}\operatorname{meas}\{t\in [T,2T]\mid f(t)\in R\}. \] Let $X=\{X(p)\mid p \text{ prime}\}$ denote a sequence of independent random variables, uniformly distributed on the complex unit circle, parameterized by the set of prime numbers. Denote the asociated random Euler product by \[ \zeta(s,X)=\prod_p (1-X(p)p^{-s})^{-1},\quad s=\sigma+it,\ \sigma>1/2. \] It's know that this product converges almost surely. Let $P(\log \zeta(s,X)\in R)$ denote the probability that the random variable $\log \zeta(s,X)$ belongs to $R$. Define the descrepancy to be \[ D_\sigma(T)=\sup_R |P_T(\log \zeta(\sigma+it)\in R)-P(\log \zeta(s,X)\in R)|, \] where the sup is taken over all rectangles whose sides are parallel to the coordinate axes. The main results of the article under review are as follows. \begin{itemize} \item For $1/2<\sigma <1$, we have $D_\sigma(T)<<(\log T)^{-\sigma},$ where the implied constant depends on $\sigma$. A similar estimate holds for $\sigma=1$. \item For $1/2<\sigma <1$ and for each sufficiently small $\epsilon>0$, we have $D_\sigma(T)=\Omega (T^{1-2\sigma-\epsilon}),$ where the implied constant depends on $\sigma,\epsilon$. \item For $a\in {\mathbb{C}}^\times$ and $1/2<\sigma_1<\sigma_2<1$, let $N_a(\sigma_1,\sigma_2,T)=\#\{(\sigma,t)\mid \sigma_1<\sigma<\sigma_2, T<t<2T, \zeta(\sigma+it)=a\}$ denote the number of $a$-points in the stated rectangle. There is a constant $c(a,\sigma_1,\sigma_2) >0$ such that \[ N_a(\sigma_1,\sigma_2,T)=c(a,\sigma_1,\sigma_2)T+ O(T\frac{\log\log T}{(\log T)^{\sigma_1/2}}). \] \item For $1/2<\sigma <1$ there is a constant $b(\sigma)>0$, such that for all $3<\tau<b(\sigma)(\log T)^{1-\sigma}(\log \log T)^{1-\frac{1}{\sigma}}$, \[ P_T(Re\{\log \zeta(\sigma+it)\}>\tau) =P(Re\{\log \zeta(s,X)\}>\tau)+\text{ error}, \] where the lower order error term is explicitly estimated. A similar estimate is obtained by replacing the real part $Re$ by the imaginary part $Im$. \end{itemize} More carefully stated versions of these results are given in the paper under review.