The sum of digits of primes in \(\mathbb Z[i]\) (Q1022349)

Recently, \textit{C. Mauduit} and \textit{J. Rivat} [Ann. Math. (2) 171, No. 3, 1591--1646 (2010; Zbl 1213.11025)] proved that the sum of digits of prime numbers in an integer base \(q \geq 2\) is equidistributed in arithmetic progressions, and that the sequence \((\alpha s_q(p))\), where \(p\) runs through the prime numbers, is equidistributed modulo 1 if and only if \(\alpha\) is an irrational real number. In this paper, similar results are proved for Gaussian integers. For any Gaussian prime \(q = -a+i\) with \(a \geq 28\) and any \(\alpha \in \mathbb{R}\) with \((a^2+2a+2) \alpha \not\in \mathbb{Z}\), it is shown that \[ \sum_{n \in \mathbb{Z}[i]: |n|^2 \leq N} \Lambda_i(n) e(\alpha s_q(n)) \ll N^{1-\sigma_q(\alpha)} \] for some constant \(\sigma_q(\alpha)\). Here, the sum-of-digits function is defined for \(n \in \mathbb{Z}[i]\) by \(s_q(n) = \sum_{j=0}^k \varepsilon_j\) where \(n = \sum_{j=0}^k \varepsilon_j q^j\) is the unique representation of \(n\) with \(\varepsilon_j \in \{0,1,\ldots,a^2\}\) and \(\varepsilon_k > 0\). Furthermore, \(e(x) = \exp(2\pi i x)\) and \(\Lambda_i(n)\) is the complex von Mangoldt function, defined by \(\Lambda_i(n) = \log |p|\) when \(n = \varepsilon p^\nu\) with \(\varepsilon\) a unit, \(p\) a Gaussian prime and \(\nu\) a positive exponent, \(\Lambda_i(n) = 0\) otherwise. >From this upper bound on exponential sums, it follows that the sequence \((\alpha s_q(p))\), running over Gaussian primes \(p\) is uniformly distributed modulo 1 if and only if \(\alpha\) is an irrational real number. Moreover, for any positive integer \(g\), the sum-of-digits function is equidistributed among the residue classes modulo \(g\), with an error term of order \(N^{1-\sigma}\) for some constant \(\sigma > 0\). The authors write that, with much more effort, one could probably cope with composite bases \(q\) whose smallest prime factor is not less than \(|q|^\alpha\), for some \(0 < \alpha < 1\). The problems with small primes share similarities with the circle problem.