5
$\begingroup$

Start with the cyclic group $G:=\mathbb{Z}/p$ of prime order $p$ and and an integer lattice $P:=\mathbb{Z}^p$. Let $G$ act on $P$ by cyclic permutation of coordinates. There is an induced action on the group ring $\mathbb{Z}[P]$ of the lattice $P$ which is isomorphic to the ring $\mathbb{Z}[\mathbf{x^{\pm 1}}]$ of Laurent polynomials in $p$ variables $\mathbf{x}=(x_1,\dots,x_p)$ over $\mathbb{Z}$.

How do I find a (hopefully) finite/minimal set of invariants which generates the ring of $G$-invariants $\mathbb{Z}[\mathbf{x^{\pm 1}}]^G$?

Since $\mathbb{Z}[\mathbf{x^{\pm 1}}]^G=\mathbb{Z}[\mathbf{x}]^G[\sigma_p^{-1}]$ where $\sigma_p=x_1\dots x_p$, the question about Invariant Polynomials under a Group Action (hidden GIT) is of some help but not much is said there about what happens over $\mathbb{Z}$.

From the Lie theory point of view this is related to the weight lattice of the root system $A_{p-1}$ where $G$ is a subgroup of the Weyl group $S_p$ and the fundamental domain I'm considering is some sort of "larger Weyl chamber".

$\endgroup$
1
  • 1
    $\begingroup$ If we take the group ring with coefficients from the ring of $p$-th cyclotomic field, then the action can be diagonalized, and invariants are readily found: then to solve your original question one has to carry out 'Galois Descent'. This is done in H W Lenstra's paper in Inventiones 1974 (in the context of Noether's problem) $\endgroup$ Jan 22, 2015 at 3:05

1 Answer 1

2
$\begingroup$

I think there is little hope that your question has a good answer.

Even over $\mathbb{Q}$, the invariants (polynomial invariants, not rational invariants, which are easier) of the cyclic group are messy. I don't see a reason why localizing by $\sigma_p$ should make things easier. Now doing the computation over $\mathbb{Z}$ adds to the difficulty since computing over $\mathbb{Z}$ means that all anomalies of the modular situation will show up.

To give evidence to my first claim, here are the numbers $r$ of minimal generators of the invariant rings $\mathbb{Q}[x_1,\ldots,x_n]^{C_n}$ for some $n$, computed by MAGMA: $n = 3$: $r = 4$; $n = 4$: $r = 7$; $n = 5$: $r = 15$, $n = 7$: $r = 48$; $n = 9$: $r = 119$; $n = 11$: $r = 348$.

$\endgroup$

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.