10
$\begingroup$

Let $G$ be a finite group and let $Rep(G)$ be its representation ring (as a group it is the free $\mathbb{Z}$-module on the irreducible complex reps). The collection of permutation representations $\mathbb{C}[\mathcal{O}]$ for $\mathcal{O}\cong G/H$ a $G$-orbit generate a $\mathbb{Z}$-subalgebra which we will denote by $Per(G) \subset Rep(G)$.

If $Per(G) = Rep(G)$ then it follows that all complex representations of $G$ are defined over $\mathbb{Z}$. To see this note that under our assumption for every representation $V$ there exists a representation $U$ defined over $\mathbb{Z}$ s.t. $U \oplus V = W$ is defined over $\mathbb{Z}$. We can now take $V_{\mathbb{Z}} = W_{\mathbb{Z}} / U_{\mathbb{Z}}$ as a $\mathbb{Z}$-form for $V$ (this is not really a precise proof, maybe a better arguemnt would be to explicitly write the projection operator which projectss onto the isotypic component of $V$ inside $\mathbb{Z}[G]$).

Is the converse true?

Question: Suppose every $\mathbb{C}$-representation of $G$ has a $\mathbb{Z}$-form, does it follow that $Per(G) = Rep(G)$? If not what's a counter example?

As an example when $G=S_n$ both of the properties are satisfied and this is in fact the only non-trivial example I know of.

$\endgroup$
2
  • 4
    $\begingroup$ Just to make sure, isn't "defined over $\mathbf{Z}$" the same as "defined over $\mathbf{Q}$", as one can always choose a $G$-stable lattice inside a $\mathbf{Q}G$-module? $\endgroup$
    – Wille Liu
    Jul 2, 2019 at 8:24
  • 2
    $\begingroup$ @WilleLiou I believe this is true. You can just take the sum of all translates of a lattice to get a $G$-stable lattice (because it is projective and projectives are free over $\mathbb{Z}$). $\endgroup$ Jul 2, 2019 at 8:32

2 Answers 2

16
$\begingroup$

The answer is no. Counterexamples include the Weyl groups of types E6, E7, and E8. For a proof that the representations of these Weyl groups are indeed all realisable over $\mathbb{Q}$ (equivalently over $\mathbb{Z}$), see M. Benard, On the Schur Indices of Characters of the Exceptional Weyl Groups, Annals of Mathematics, Vol. 94, No. 1 (Jul., 1971), pp. 89-107 (MSN). For a proof of the statement that $Per(G)\neq Rep(G)$ for these groups, see D. Kletzing, Structure and Representations of Q-Groups, Lecture Notes in Mathematics 1084 (MSN).

$\endgroup$
11
$\begingroup$

The answer is no. For a discussion and counter example, see (for example)
MR3451397 Bartel, Alex; Dokchitser, Tim Rational representations and permutation representations of finite groups. Math. Ann. 364 (2016), no. 1-2, 539–558. https://arxiv.org/abs/1405.6616

$\endgroup$
5
  • $\begingroup$ Indeed, the dihedral group of order 8 gives already a counter example. $\endgroup$
    – Wille Liu
    Jul 2, 2019 at 9:07
  • 3
    $\begingroup$ @WilleLiou Actually, that's not true. As mentioned in the introduction of the above paper, there is no counter example when $G$ is a $p$-group. $\endgroup$ Jul 2, 2019 at 9:17
  • 2
    $\begingroup$ @HenriJohnston If I understand correctly the paper shows that it is possible for $Rep_{\mathbb{Q}}(G) / Perm(G)$ to be non-trivial in general (i.e. the quotient of the ring of virtual rational representations by $Perm(G)$). Though it doesn't assume (as in my question) that $Rep_{\mathbb{Q}}(G) = Rep_{\mathbb{C}}(G)$. I tried to look there whether there's a counter example to this as well but I didn't find it. No doubt its there and I just missed it... $\endgroup$ Jul 2, 2019 at 9:57
  • $\begingroup$ @SaalHardali sorry, I missed the $Rep_{\mathbb{Q}}(G) = Rep_{\mathbb{C}}(G)$ requirement. Of course, Alex's answer above fully answers your question. $\endgroup$ Jul 2, 2019 at 17:31
  • $\begingroup$ @HenriJohnston No worries. I'm happy that you posted this answer if only for this very nice and readable paper. I would accept both answers if I could. $\endgroup$ Jul 2, 2019 at 17:33

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.