It is well known that for abelian number fields, the factorization of its Dedekind zeta function goes like this:

$$\zeta_K(s)=\zeta(s)\prod_{\chi \neq 1} L(s,\chi)$$

with the Dirichlet characters distinct and primitive.

If $K$ is non-abelian but Galois, we instead have (by Aramata-Brauer):

$$\zeta_K(s)=\zeta(s)\prod_{\rho \neq 1} L(s,\rho)$$

with the representations non-trivial and irreducible, although in this case we don't know unconditionally that that's a factorization of irreducible L-functions. In the non-Galois case, factors might appear with arbitrary integer powers. But let's forget about that for a moment.

In this answer, Kevin Dong mentions an explicit factorization of the zeta of $\mathbb{Z}[\sqrt[3]{2}]$ in terms of modular forms. It is very nice but not quite suprising: Artin L-functions are expected (and known in some cases) to always be automorphic.

  • I'm interested on the proof for $\mathbb{Z}[\sqrt[3]{2}]$ (I haven't been able to find a reference for it), and any other reference for known case of such a factorization (this is, not a factorization on terms of Artin L-function, but of automorphic ones).

  • I'd also want to know what the conjectures are (on the automorphic side) for what the factorization looks like for an arbitrary non-Galois Dedekind zeta function.

Any other information around those issues might be of help, but not generally about Langlands or the Artin conjecture.

  • 4
    $\begingroup$ The rule of thumb is this: the $L$-function of an irreducible $n$-dimensional Galois representation equals (conjecturally) the $L$-function of an (algebraic) cuspidal representation of $\mathrm{GL}_n$ over $\mathbb{Q}$. That is, in the case of $K$ Galois, only Dirichlet $L$-functions and modular $L$-functions appear in the factorization of $\zeta_K(s)$ if and only if all irreducible representations of $\mathrm{Gal}(K/\mathbb{Q})$ are at most two-dimensional. Note that modular $L$-functions include the $L$-functions of Maass forms. $\endgroup$
    – GH from MO
    Jul 31, 2015 at 0:09
  • 3
    $\begingroup$ @KevinVentullo: Actually some researchers (like myself) also include Maass forms among modular forms, and distinguish "the older modular forms" by saying "holomorphic modular forms". $\endgroup$
    – GH from MO
    Jul 31, 2015 at 0:17
  • 2
    $\begingroup$ Okay fair enough, I just wanted to avoid confusion for the OP $\endgroup$ Jul 31, 2015 at 0:18
  • 2
    $\begingroup$ @Myshkin: By standard conjectures, the Artin $L$-functions are primitive, so yes, the known instances of factoring a Dedekind zeta are precisely when the Artin $L$-functions occurring in the factorization are known to satisfy Artin's conjecture (i.e. when they are holomorphic). $\endgroup$
    – GH from MO
    Jul 31, 2015 at 0:56
  • 1
    $\begingroup$ @Myshkin: The "only if" part follows from the first sentence (i.e. the strong Artin conjecture) and the fact that $L$-functions of cuspidal representations are expected to be primitive. For example, if (conjecturally) a $\mathrm{GL}_3$ cuspidal $L$-function is present, you cannot factor it into $\mathrm{GL}_1$ and $\mathrm{GL}_2$ $L$-functions. For your other comment: I don't understand what you mean by a concrete factorization. Artin $L$-functions are pretty concrete in my opinion. $\endgroup$
    – GH from MO
    Jul 31, 2015 at 1:47

1 Answer 1


if you know theta series, you write $L(s,\rho)= \sum_{n>0 } a_n n^{-s}$ and you have ; $\sum_{n>0} a_nq^n = 1/2(\Theta_{1,0,27}-\Theta_{4,2,7})$. Where : $$ \Theta_{a,b,c} = \sum_{(x,y) \in \mathbb{Z}^2} q^{ax^2+bxy+cy^2} $$

For diedral odd representation all is explicit.

  • $\begingroup$ For non-abelian cubic extension $K=Q[x]/(f), D = disc(f) $( $ < 0$),$ F= Q(\sqrt{D}), L = KF$ then $\zeta_K(s)=\zeta(s)L(s,\rho : Gal(L/Q)\to GL_2)$ where $\rho$ is the dihedral representation of $S_3 = Gal(L/Q)$ and by class field theory or by CM on the elliptic curve with CM by an order of $O_F$ such that $F(j(E))=L$ then $L(s,\rho,Gal(L/Q)\to GL_2)=L(s,\psi : Gal(L/F)\to GL_1)=L(s,\psi,F)$ $\endgroup$
    – reuns
    Feb 3, 2019 at 6:11
  • $\begingroup$ where the latter is an Hecke L-function, corresponding to an automorphic representation of $GL_1(F)\setminus GL_2(\mathbb{A}_F)$. In addition $L(s,\psi,F)) = L(s,f)$ with $f(z) = \sum_{a \in O_F } \psi(a) e^{2i \pi |a|^2 z} \in S_{1+2r}(\Gamma_1(disc(O_F) m))$, $m$ the conductor of $\psi$ (the series you wrote, with one quadratic form by ideal class of $\mathbb{Z}+m O_F$) and $|\psi(a)| = |a|^r$ which gives an automorphic representation of $GL_2(\mathbb{Q})\setminus GL_2(\mathbb{A_Q})$. $\endgroup$
    – reuns
    Feb 3, 2019 at 8:51
  • $\begingroup$ This series $\sum_{n=1}^\infty a_n q^n$ also has the eta-product representation $\eta(q^6) \eta(q^{18}) = q \prod_{m=1}^\infty (1-q^{6m}) (1-q^{18m})$. $\endgroup$ Feb 22, 2021 at 1:14

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.