Let $A$ be an associative algebra over a filed $k$.

Q) What are the condition we can impose on $A$ such that there exists a $G$ such that $A=k[G]$, the group algebra generated by $G$?

I am particularly interested in the following cases:

1) When $k=\mathbb{Q},\mathbb{R}$ or $\mathbb{C}$.

2) $A=M_n(k),$ the matrix algebra or a subalgebra of matrix algebra.

PS: I am not sure whether this question is of research level or not. If anybody thinks that this is not proper here please give the references and then vote to close. I have searched it in general but could not find any answer. Thanks in advance.


If $k$ has characteristic zero and $A$ is finite-dimensional, by Maschke's theorem it must be semisimple. If $k$ is in addition algebraically algebraically closed and $A \cong k[G]$, then $A \cong \prod_i M_{n_i}(k)$ where $n_i$ are the dimensions of the irreducible representations of $G$. So in this case the question reduces to the following purely group-theoretic question:

When is a multiset $n_i$ of positive integers the multiset of dimensions of the irreducible representations of a finite group over $k$?

Some miscellaneous necessary conditions are that

  • one of the $n_i$ must be equal to $1$ (the trivial representation),
  • each $n_i$ must divide $\dim A = |G| = \sum n_i^2$,
  • the number of $n_i$ equal to $1$ (the size of the abelianization of $G$) must also divide $\dim A$,
  • if $\dim A$ is prime (so $G$ must be cyclic), then each $n_i = 1$.

The first condition rules out $M_n(k)$ for $n \ge 2$. The second condition rules out lots of examples, such as $k \times M_2(k)$. The third condition rules out lots more examples, such as $k \times k \times M_3(k)$. And so forth. I can't imagine there are any useful sufficient conditions; there are a lot of finite groups and they can be very complicated.

| cite | improve this answer | |
  • $\begingroup$ @yuan Thanks for the answer. You gave the perfect answer I was looking for. Just one question: Does the infinite dimension case has been studied or can be derived from some results? $\endgroup$ – Cusp May 14 '15 at 6:02
  • 1
    $\begingroup$ @Cusp: group rings of infinite groups seem pretty difficult to understand. There are some big open problems in this area: see, for example, en.wikipedia.org/wiki/Kaplansky%27s_conjecture. $\endgroup$ – Qiaochu Yuan May 14 '15 at 6:05
  • $\begingroup$ It might, however, be interesting to notice that $M_n(k)$ can be realised as a groupoid algebra. $\endgroup$ – Mateusz Wasilewski May 14 '15 at 7:13
  • $\begingroup$ @Wasilewski Can you please elaborate your statement? $\endgroup$ – Cusp May 14 '15 at 16:04
  • $\begingroup$ If you have a groupoid of pairs $(i,j)$, where $i,j \leqslant n$, endowed with the partial composition operation $(i,j)\cdot (j,k) = (i,k)$ and not defined for pairs $(i,j)$ and $(k,l)$ if $j\neq k$, then the groupoid algebra is isomorphic to $M_n(k)$. This is not surprising because $(i,j)$ look like matrix units. $\endgroup$ – Mateusz Wasilewski May 14 '15 at 16:27

If you take a solvable finite dimensional associative algebra $A$ with 1 having a separable radical factor structure, then, by the Wedderburn-Malcev-Theorem, there exist an abelian subalgebra $T$ with $A=rad(A)\oplus T$. The unit group $E(A)$ is the semidirect product of $1+rad(A)$ and $E(T)$. You can show that $\langle E(A) \rangle _K=A$ holds, if the the underlying field has more than two elements. Therfore you get a surjection from $KE(A)$ to $A=\langle E(A) \rangle _K$. So here we have another connection but not equality. (solvable is defined that $A/rad(A)$ is commutative).

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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