On sums of squares of primes. II (Q983298)

The authors improve their results given in the previous paper [Part I, Math. Proc. Camb. Philos. Soc. 140, No. 1, 1--13 (2006; Zbl 1114.11079)], which has a serious oversight in the proofs. They prove, with \[ A_3:=\{ n\in \mathbb N : n\equiv 3\pmod {24}, n\not \equiv 0\pmod 5\} \] and \[ A_4:=\{ n\in \mathbb N : n\equiv 4\pmod {24}\}, \] that the exceptional sets-cardinalities, for \(j=3,4\), \[ E_j(N):=\left| \{ n\in A_j : n\leq N, n\neq p_1^2+\cdots+p_j^2, \forall p_u\}\right|, \] where the letters \(p_u\) denote primes, are bounded as \(E_3(N)\ll N^{17/20+\varepsilon}\) and \(E_4(N)\ll N^{7/20+\varepsilon}\); also, another bound is given for the cardinality of the exceptional set forming the natural numbers up to \(N\) (which have necessary congruence conditions) that are not representable as \(p+p_1^2+p_2^2\), i.e. \(E(N)\ll N^{7/20+\varepsilon}\). The methods applied (an embellishment of the ones in Part I, quoted above) are within the circle method and, also, furnish better estimates for the exponential sums involved; inasmuch employing sieve methods, these do not give asymptotic formulae for the number of representations, but only lower bounds.