Explicit multiplicative relations between Gauss sums (Q996274)

For an integer \(e>2\) and a prime \(p\equiv 1\bmod e)\), we consider the multiplicative character \(\chi: \gamma\in({\mathbb Z}/p{\mathbb Z})^\times\mapsto \zeta_e\in {\mathbb C}\) where \(\gamma\) is a primitive root modulo \(p\) and \(\zeta_e\) is a primitive \(e\)-th root of unity. If \(\zeta_p\) is a primitive \(p\)-th root of unity, we have the associated Gauss sums (indexed by \(a\in{\mathbb Z}\)) \(\tau(a)=\sum_{\alpha\bmod p}{\chi^a(\alpha)\zeta_p^\alpha}\). This paper deals with the problem of exhibiting multiplicative relations -- coming neither from the Hasse-Davenport relation nor from the norm relation \(| \tau(a)| =\sqrt{p}| \) -- between the sums \(\tau(a)\) as \(a\) varies. Such relations -- known as \textit{sign ambiguities} -- are known to exist (see \textit{K. Yamamoto} [Symp. Math. 15, Inf. Teor., Strutt. Corpi Algebr., Convegni 1973, 427--440 (1975; Zbl 0319.10044)]) though no general explicit formula had ever been proven so far. In this paper B. J. Murray produces such formulae, for infinitely many cases (i.e. for infinitely many values of \(e\)). The method used is very elementary but combines many clever tricks. First, the author considers carefully different types of sums of integers indexed by the group of biquadratic residues modulo \(e\) and shows these sums are closely related to the order \(h_K\) of the class group of the quadratic field \(K={\mathbb Q}(\sqrt{-e})\) (as a corollary -- and with the assumptions the author needs on \(e\) -- the congruence of \(h_K\) modulo \(8\) is determined). The study of those sums yields an equality between ideals in \({\mathbb Q}(\zeta_e)\) which is then translated as an equality between algebraic numbers (up to multiplication by a unit of \({\mathbb Q}(\zeta_e)\)). Using a criterion of Yamamoto (see the above reference), that relation is proven to come neither from the Hasse-Davenport formula nor from the norm formula. Finally the author resolves the remaining ambiguity (i.e. he determines the value of the unit involved in his relation) using elementary Galois correspondence and his intermediate results on \(h_K\). Reviewer's remark: One misprint page 88: in the first paragraph of Section 2, one should read \(\tau(a) \overline{\tau(a)}=p\) in stead of \(\tau(a) \overline{\tau(a)}= \chi_a(-1)p\).