There is a wide litterature for the classical Gauss sums. For $\chi$ a primitive Dirichlet character modulo $N$, it is given by $$\tau(\chi) = \sum_{n \text{ mod } N} \chi(n) \exp(2i\pi n/N).$$

An interesting fact is that primitive Dirichlet characters are essentially their own Fourier transform, up to this associated Gauss sum: $$\sum_{n \text{ mod } N} \chi(n) \exp(2i\pi nm/N) = \bar{\chi} (m) \tau(\chi). \qquad (\star)$$

I am interested in similar properties in general number fields $F$. Let $\omega$ be a nonzero element of $F$ and write $\mathfrak{a}$ for its denominator. Hecke defined some analogous Gauss sums by $$C(\omega) = \sum_{n \text{ mod } \mathfrak{a}} \exp(\mathrm{tr}(n^2 \omega)).$$

(here $n \text{ mod } \mathfrak{a}$ means that $n$ is an integer ideal in $\mathfrak{o}_F/\mathfrak{a}\mathfrak{o}_F$, where $\mathfrak{o}_F$ is the ring of integers of $F$). I have not found anything concerning the analogous Gauss sums as above, namely for a finite order character $\chi$ on integer ideals which is a character modulo $\mathfrak{a}$, $$\tau(\omega, \chi) = \sum_{n \text{ mod } \mathfrak{a}} \chi(n)\exp(\mathrm{tr}(n^2 \omega)).$$

Is there anything of this kind in the litterature? In particular, do we have an analogous result to $(\star)$?

  • $\begingroup$ Do you really want $n$ to be an ideal? You can find other treatments of Gauss sums over number fields and local fields in some articles in the 1977 Durham conference proceedings "Algebraic Number Fields" edited by Frohlich (not to be confused with the book by Cassels & Frohlich). See the contributions by Martinet on Artin $L$-functions and by Tate on local constants. $\endgroup$
    – KConrad
    Commented Nov 9, 2018 at 3:32
  • $\begingroup$ @KConrad Thanks for the reference, I will try to find it. However I am truly interested in sums over ideals, neither over norms nor over elements of a certain ideal. I know there are generalizations of this kind (for instance Evans or Skoruppa for norms, ) with sums over ideals but using standard Dirichlet characters of the norm of the ideal, not characters (of the ideal class group). $\endgroup$ Commented Nov 9, 2018 at 4:10

1 Answer 1


Hecke Gauss sums are quadratic Gauss sums, and (*) is correct for quadratic characters (Hecke, Satz 155). Hecke Gauss sums were studied later by

  • Siegel, Über das quadratische Reziprozitätsgesetz in algebraischen Zahlkörpern, Nachr. Akad. Wiss. Göttingen, Math. Phys. Kl. II (1960), 1-16; Ges. Abh. 3 (1966), 334-349
  • Shiratani, On the decomposition laws of rational primes in certain class 2 extensions, Investigations in number theory, Advances in Pure Math. 13 (1988), 345-411
  • Shiratani, On the Gauss-Hecke-sums, J. Math. Soc. Japan 16 (1964), 32-38
  • Boylan, Skoruppa, A quick proof of reciprocity for Hecke Gauss sums, J. Number Theory 133 (2013), 110-114.
  • $\begingroup$ Thanks for your answer. I already bumped into some of these references and I confess that the question is in fact motivated by the fact that Boylan and Skoruppa consider the Hecke-Gauss sums as sums over characters. Is there anything known analogously to (*) for this case? $\endgroup$ Commented Nov 10, 2018 at 9:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.