On Selberg's central limit theorem for Dirichlet \(L\)-functions (Q2219653)

It was shown by \textit{A. Selberg} [Arch. Math. Naturvid. 48, No. 5, 89--155 (1946; Zbl 0061.08402)] that \(\log\zeta(\tfrac12+it)\) follows a 2-dimensional Gaussian distribution, with mean 0 and variance \(\tfrac12\log\log|t|\). A new approach, which handles the real part \(\log|\zeta(\tfrac12+it)|\), was recently given by \textit{M. Radziwiłł} and \textit{K. Soundararajan} [Enseign. Math. (2) 63, No. 1--2, 1--19 (2017; Zbl 1432.11117)]. This new approach avoids the need for the delicate zero-density estimate used by Selberg. The present paper applies the method of Radziwiłł and Soundararajan to Dirichlet \(L\)-functions, and shows indeed that the logarithms of distinct \(L\)-functions are independent. Specifically, suppose that \(\chi_1,\ldots,\chi_N\) are distinct primitive Dirichlet characters. Then when \(T\to\infty\) and \(t\) runs over the interval \([T,2T]\), the vector \[ \left(\log\left|L(\tfrac12+it,\chi_1)\right|,\ldots, \log\left|L(\tfrac12+it,\chi_1N)\right|\right) \] follows an approximate \(N\)-variate normal distribution with mean zero and covariance matrix \(\tfrac12(\log\log T)\mathbf{I}_N\).