Are there any examples of finite-dimensional complex non-semisimple non-commutative symmetric Frobenius algebras? Or can one show that none exist?

I went through this list of all complex associative algebras up to dimension $4$ and couldn't find any non-commutative non-semisimple ones that can be equipped with a corresponding linear form to make them into a symmetric Frobenius algebra. Are there examples in higher dimensions?


2 Answers 2


Given any finite dimensional algebra $A$, consider the linear dual $\hat{A}= \hom(A, k)$ as an $A$-$A$-bimodule. Then $R = A \oplus \hat{A}$ may be equipped with an algebra structure as follows:

$$(a, x) \cdot (b,y) = (ab, x \cdot b + a \cdot y)$$

for $a,b \in A$ and $x,y \in \hat{A}$. The algebra $R$ has a natural symmetric Frobenius algebra structure. So every finite dimensional (possibly non-commutative) algebra embeds into a symmetric Frobenius algebra.

  • 1
    $\begingroup$ Is it obvious that there are (non-commutative) examples for which the resulting symmetric Frobenius algebra is non-semisimple? $\endgroup$
    – Andi Bauer
    Mar 13, 2021 at 15:40
  • 3
    $\begingroup$ @AndiBauer yes it is, since $\hat{A}$ is then a 2-sided ideal with zero law. So the resulting $R=R_A$ is not semisimple unless $A=0$ (and not commutative as soon as $A$ is not). $\endgroup$
    – YCor
    Mar 13, 2021 at 15:44
  • 2
    $\begingroup$ The bilinear form is $\langle (a,f), (b,g) \rangle = f(b) + g(a)$. $\endgroup$ Mar 15, 2021 at 18:22
  • 1
    $\begingroup$ @AndiBauer The standard actions of $A,A$ on $Hom(M,k)$ would be $(r\cdot f)(m)=f(mr)$ and $(f\cdot r)(m)=f(rm)$, which is my estimate at what is intended. $\endgroup$
    – rschwieb
    Mar 16, 2021 at 13:07
  • 1
    $\begingroup$ @AndiBauer You said "algebras". I assumed you meant unital or would have said otherwise. Of course unital algebras are also not-necessarily unital algebras, so this still gives examples asked for in the OP. If A is unital, then the bilinear form is induced by the counit $\epsilon(a,f) = f(1)$. If $A$ is non-unital, then I guess there is no counit, but the bilinear form still makes sense and is symmetric and non-degenerate. $\endgroup$ Mar 17, 2021 at 13:11

I just realize that my question is actually rather trivial the way I posed it: There exist non-commutative semisimple examples (with the $2\times 2$ matrix algebra being the smallest example). There also exist commutative non-semisimple examples, the smallest one given by $$e_0\cdot e_0 = e_0,\quad e_0\cdot e_1=e_1,\\ e_1\cdot e_0=e_1,\quad \mu(e_1)=1\;,$$ where $\mu$ is the linear form generating the symmetric non-degenerate pairing and all other entries $0$.

All one needs to do is consider direct sums or tensor products of a non-commutative (but semisimple) and a non-semisimple (but commutative) example, and it will be both non-commutative and non-semisimple. The smallest such example is of dimension $6$.

I guess a more sensible but a bit artificial question to ask would have been whether there are examples that do not come from the tensor product or direct sum of a commutative and a semi-simple example.


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.