MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Given a class function $f: G \to \mathbb Q$, where $G$ is a finite abelian group, is there an easy way to decide whether $f$ is an element of the rational representation ring $R_{\mathbb Q}(G)$, i.e. whether $f$ is a virtual character of some representations of $G$?

If it makes things easier you could also assume that $G$ is a $p$-group and the class function has values in $\mathbb Z$.

share|cite|improve this question
up vote 3 down vote accepted

This essentially boils down to the case of a cyclic group. For a cyclic group of order n, the irreducible representations correspond to the action on $\mathbb Q[\omega_d]$ where $\omega_d$ is a primitive $d^{th}$-root of unity where $d$ divides $n$. So one can easily produce the rational character table and check if your function is a non-negative linear combination of these irreducible characters.

Added: You can now use the orthogonality relations over $\mathbb Q$ to see if something is a character of a rep. One knows the endomorphism algebra of $\mathbb Q[\omega_d]$ as a module over the cyclic group $G$ is $\mathbb Q[\omega_d]$ which has degree $\phi(d)$. So the orthogonality relations (valid over any field of characteristic 0, but in the non-algebraically closed case involves the dimensions of the endomorphism division algebras) can be used to decompose class functions in terms of irreducible characters.

Answer 2 Alternatively, one can build the character table of a finite abelian group $G$ over $\mathbb Q$ in the following way (without reducing to the case of a cyclic group). The Schur index of a degree one character is $1$. So the irreducible characters for $G$ are obtained as follows. Take a complex character $\chi$ and let $\Gamma(\chi)=Gal(\mathbb Q[\chi]:\mathbb Q)$ where $\mathbb Q[\chi]$ is the field generated by the values of $\chi$. Then the sum of the orbit of $\chi$ under $\Gamma$ is a $\mathbb Q$-irreducible character and all such characters are obtained in this fashion. So this describes an orthogonal basis for class functions over $\mathbb Q$ and hence all class functions.

supplement This may be off, but I think a rational-valued class function $f$ is a rational character iff for each complex character $\chi$, one has the inner product $\langle f,\chi\rangle$ is an integer and this integer is constant on the orbit of $\chi$ under $\Gamma(\chi)$.

share|cite|improve this answer
Could you explain a little more why "this essentially boils down to the case of cyclic groups"? – Dimitrios Sep 29 '11 at 10:29
Every finite abelian group is a direct product of cyclic groups. A representation of a direct product is a tensor product of representations of the factors. So the characters of an abelian group over $\mathbb Q$ will be products of characters of cyclic groups over $\mathbb Q$. So once you compute the character table of a cyclic group, as I explained how to do above, then you can compute the character table of any abelian group. – Benjamin Steinberg Sep 29 '11 at 12:33
The formula $R_{\mathbb Q}(G_1\times G_2)\cong R_{\mathbb Q}(G_1)\otimes R_{\mathbb Q}(G_2)$ only holds if the orders of the (finite) groups $G_1$ and $G_2$ are coprime. So especially in the case where $G$ is a $p$-group you can't reduce the problem to cyclic subgroups. I might be wrong though or I didn't get something right. – Dimitrios Sep 29 '11 at 13:11
One more question concerning your example: $\mathbb Q[\omega_3]\otimes\mathbb Q[\omega_3]\cong \mathbb Q[\omega_3]\times \mathbb Q[\omega_3]$, thus it is not irreducible as an $\mathbb Q(G_1\times G_2)$-module, or is it? – Dimitrios Sep 29 '11 at 16:12
Your are right of course and what I wrote was dumb. The nontrivial irreducible representations of $(Z/p)^n$ over $\mathbb Q$ all have degree $p-1$. Indeed, this is true for $Z/p$. Now there are (p^n-1)/(p-1) distinct homomorphisms from $(Z/p)^n$ to $Z/p$. So this gives $(p^n-1)/(p-1)$ distinct irreducible reps of degree $p-1$. Thus in the regular representation this accounts for all nontrivial irreducible reps. I think my second answer using Schur index is correct. – Benjamin Steinberg Oct 4 '11 at 16:05

A necessary criterion is that, if $g \in G$ has order $m$, and $k$ is relatively prime to $m$, then $f(g) = f(g^k)$.

Proof: Let $\rho$ be a rational representation. Let the eigenvalues of $\rho(g)$ be $\omega_1$, $\omega_2$, ..., $\omega_N$. Then $f(g) = \sum \omega_i$ and $f(g^k) = \sum \omega_i^k$.

But the $\omega_i$ are all $m$-th roots of unity, and there is a Galois symmetry $\sigma$ of $\mathbb{Q}(e^{2 \pi i/m})$ which acts on $m$-th roots of unity by $\omega \mapsto \omega^k$. So $f(g^k)$ is the image of $f(g)$ under this symmetry. Since $f(g)$ is in $\mathbb{Q}$, this shows that $f(g) = f(g^k)$. QED

One can show that any rational class function as above is a $\mathbb{Q}$-linear combination of rational characters. If you want to know about $\mathbb{Z}$-linear combinations, that sounds harder; I don't know a simple rule.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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