2
$\begingroup$

Let $G$ be a compact topological group with normalized Haar measure $\mu$.

Is there an effective isometric action of $G$ on some $\mathbb{R}^n$ such that the following map would be a non-irreducible polynomial in $\mathbb{R}[x_1,x_2,\ldots,x_n]$?

$$P(X)=\int_G \langle X, \; g.X\rangle,\qquad X\in \mathbb{R}^n$$

A motivational example is the permutational action of the finite group $G$ generated by cyclic permutation $(1,2,\ldots,n)$ on $\mathbb{R}^n$. For this action the above map $P$ has the formula

$$P(x_1,x_2,\ldots,x_n)=\frac{(x_1+x_2+\ldots+x_n)^2}{n}$$

My initial motivation was the following:

I encountered the inquality $n\sum_{i=1}^n x_i^2 \geq (\sum_{=1}^n x_i)^2$. To prove this I realized that the right hand side of this inequality has the expansion $\sum_{i=0}^n <X, g^i.X>$ where $g$ is the cyclic permutation $g=<2,3,\ldots,n,1>$. Then I applied Cauchy Schwartz inequality. So this leads me to be interested in the term $\int_G <X, gX>$ in present of a group action. According to the comment of Prof. Valette I realized that this integral is $|Q(x)|^2$ where $Q$ is projection on the space of fixed points of $G$.

My reason to emphasis on effectiveness was that: for trivial action we get the irreducible polynomial. Now thanks to the comment of Prof. Valette I realize that the only non irreducible case is when the space of $G$ fixed vectors is a 1 dimensional space.

If the answer to the above question is yes one may associate an invariant to a compact in particular finite group $G$ equal to the minimum of all possible $n$ with the above mentioned effective isometric action of $G$ on $\mathbb{R}^n$.

What is this function $P$ when $G$ is the whole group $O(n)$?

What is the ideal generated by all polynomials $P$ obtained in this way via finit subgroups of $O(n)$?

$\endgroup$
21
  • 2
    $\begingroup$ By very simple reasons there is not even an effective action. Let $A$ be a non-trivial finite group and let $K=A^I$ for some set $I$. The cardinality of $K$ can be made arbitrarily large, i.e., larger than the set of permutations on $\mathbb{R}^n$. $\endgroup$
    – user473423
    Commented Feb 17, 2023 at 15:26
  • 1
    $\begingroup$ No. As soon as the cardinality is large enough the group homomorphism has a non-trivial kernel. $\endgroup$
    – user473423
    Commented Feb 17, 2023 at 15:49
  • 3
    $\begingroup$ The operator $\int_G gd\mu(g)$ is the orthogonal projection $Q$ on the space of G-fixed vectors. Hence $P(X)=\|Q(X)\|^2$ is à quadratic form. $\endgroup$ Commented Feb 17, 2023 at 19:28
  • 2
    $\begingroup$ No there are cases without an effective action. Effectivity means that the group homomorphism $G\to Per(\mathbb{R}^n)$ is injective. Now if the cardinality of $G$ exceeds the one of $Per(\mathbb{R}^n)$, there cannot be an effective action. $\endgroup$
    – user473423
    Commented Feb 18, 2023 at 14:42
  • 3
    $\begingroup$ @Echo I've never seen this notation (while often encountering the notation $\mathrm{Sym}(X)$ or $S_X$, or $\mathfrak{S}_X$). Are there papers where this notation $\mathrm{Per}(X)$ is used for the group of permutations of $X$? (I'm just curious since I tried to have an overview of them.) $\endgroup$
    – YCor
    Commented Feb 20, 2023 at 10:54

0

You must log in to answer this question.