Determinations of Jacobsthal sums (Q788760)

For a positive integer n, a prime \(p\equiv 1 (mod 2n)\) and an integer a with \((a,p)=1\), the Jacobsthal sum \(\phi_ n(a)\) is defined by \(\phi_ n(a):=\sum^{p}_{m=1}(m/p)((m^ n+a)/p),\) where \((\quad /p)\) denotes Legendre's symbol. \textit{E. Jacobsthal}, with whom this whole theory started [J. Reine Angew. Math. 132, 238-245 (1907)] introduced and evaluated \(\phi_ 2(a)\) in order to give explicit values for x and y in \(p=x^ 2+y^ 2\), \(p\equiv 1 (mod 4)\). More precisely: \(\phi_ 2(a)=\pm 2 (2/p) a_ 4\) if \((a/p)=1\) and \(\phi_ 2(a)=\pm 2 | b_ 4|\) if \((a/p)=-1\). Here \(a_ 4\) and \(b_ 4\) are integers with \(a_ 4^ 2+b_ 4^ 2=p\), \(a_ 4\equiv -(2/p) (mod 4)\). In the case \((a/p)=1\) we have the \(+\) sign if and only if a is also a quartic residue mod p. The sums \(\phi_ n(a)\) with \(n=2,3,4,6,10\) and 12 have been extensively studied in recent years, in particular by \textit{B. C. Berndt} and the present author in a long paper entitled ''Sums of Gauss, Jacobi, and Jacobsthal'' [J. Number Theory 11, 349-398 (1979; Zbl 0412.10027)]. In the formulae in the latter paper there occur sign ambiguities similar to the one in the case \(n=2\), \((a/p)=-1\). The present paper removes these sign ambiguities by deriving for those \(\phi_ n(a)\) whose sign is not yet determined, certain congruences. E.g., for \(n=2\), \((a/p)=-1\) one has \(\phi_ 2(-a)\equiv 2 a_ 4 a^ f (mod p),\) where f is given by \(p=2nf+1\). This removes the sign ambiguity since \(\phi_ n(a)\) and \(\phi_ n(-a)\) are connected by the relation \(\phi_ n(a)=\phi_ n(-a) (-1)^{fn+f}.\) Every proof is linked up with the proof of the corresponding theorem in the 1979-paper of which it resolves the sign ambiguity. For the cases \(n=2,3\) and 4 other authors have given results of a similar kind but the present attack by way of Jacobi sums is new and works for all mentioned values of n.