On a sum of three prime squares (Q5928014)

Let \({\mathcal A}\) denote the set of positive integers \(n\) such that \(n\equiv\pmod{24}\) and \(n\not\equiv 0\pmod 5\). It is conjectured that every sufficiently large integer \(n\in{\mathcal A}\) can be written in the form \(n=p^2_1+ p^2_2+ p^2_3\), with \(p_1,p_2,p_3\) primes. Let \(E(x)\) be the number of integers \(n\leq x\) belonging to the set \({\mathcal A}\) which are not the sum of the squares of three primes. It is shown that \[ E(x)\ll_\varepsilon x^{151/160+ \varepsilon} \] for any \(\varepsilon>0\). Unfortunately there appears to be a misprint in the statement of the second result, but the reviewer believes the correct version to be that under the Generalized Riemann Hypothesis one has \[ E(x) \ll_\varepsilon x^{7/8+ \varepsilon}. \] Previously \textit{M.-C. Leung} and \textit{M.-C. Liu} [Monatsh. Math 115, 133-169 (1993; Zbl 0779.11045)] had proved unconditionally the existence of a positive \(\delta\) such that \(E(x)\ll x^{1- \delta}\). The most interesting feature of the argument is that neither hypothetical exceptional real zeros of quadratic \(L\)-functions, nor explicit zero-free regions for \(L\)-functions, play any special role. The tools used are the circle method, and bounds for exponential sums over primes. The latter require mean value estimates and zero-density estimates for \(L\)-functions.