On the number of primitive representations of integers as sums of squares (Q2369900)

For \(k,n\in \mathbb Z^+\) denote by \(r_k(n)\) the number of representations of \(n\) at a sum of \(k\) squares, and by \(r_k^p(n)\) the number of primitive representations. (The latter are those in which the \(k\) squares have gcd\,1.) The authors present very nice formulas for \(r_k^p(n)\), for all \(N\in\mathbb Z^+\), when \(2\leq k\leq 8\). For \(9\leq k\leq 12\), they derive formulas for \(r^p_k(n)\) under suitable congruence restrictions on \(n\). For even \(k\) the proof depends upon identities for the \(k\)th power of Jacobi's theta function \(\varphi(q) = \sum^\infty_{-\infty}q^{n^2}\), \(|q|<1\), due to Jacobi \((k=2,4,6,8)\) and Glaisher \((k=10,12)\). For odd \(k\) the relevant identities for \(\varphi^k\) have been proved by Hirschhorn and Sellers \((k=3)\) and Cooper \((k=5,7,9,11)\). The proof proceeds by invoking these identities and applying a relationship between \(r^p_k(n)\) and \(r_k(n)\) obtained from the identity \(r_k(n)=\sum_{d^2|n}r^p_k (n/d^2)\) and Möbius inversion. (See display (1,8), p. 8, of the article for the explicit representation of \(r^p_k(n)\) in terms of \(r_k(n)\).)