Let $A$ be a $\mathbb{C}$-algebra.

For finite dimensional representation $\rho:A \rightarrow \rm{End}(V)$, the trace $\chi_\rho$ of $\rho$ is defined by $\chi_\rho(a):=\rm{Tr}(\rho(a))$ for $a \in A$.

It is true that if two finite dimensional representations $\rho,\sigma$ are isomorphic then $\chi_\rho =\chi_\sigma$? Now, if $\chi_\rho = \chi_\sigma$, are $\rho$ and $\sigma$ isomorphic?


Not in general. For instance, let $A=\mathbb{C}[x]/(x^2)$. Consider two different representations of $A$ on $\mathbb{C}^2$: $\rho$, which sends $x$ to $\begin{pmatrix} 0 & 1 \\ 0 & 0\end{pmatrix}$, and $\sigma$, which sends $x$ to $0$. For any $a\in A$, $\rho(a)$ and $\sigma(a)$ have the same diagonal entries, and hence the same trace. But $\rho$ and $\sigma$ are not isomorphic.

  • $\begingroup$ Oh! it's simple counter example. Thank you. $\endgroup$ – H.Kakuhama Sep 27 '16 at 9:45

That the answer to your question is „no” has been shown in Eric Wofsey's answer. But there is a little bit more that can be said.

Let $A$ be a finite dimensional algebra over a field $\mathbb{k}$ of characteristic zero, and $M_1, M_2$ finitely generated $A$-modules with corresponding characters $\chi_1, \chi_2$. Then $\chi_1 = \chi_2$ if and only if $M_1$ and $M_2$ have the same composition factors, counted with multiplicities. Alternatively, $[M_1] = [M_2]$ as elements of the Grothendieck group $G_0(A)$.

In the counterexample of Eric Wofsey, both modules have the composition factor $\mathbb{C}$ (on which $x$ acts trivially) with multiplicity 2, so their characters have to coincide.

Details can be found in chapter 3 of [Lam - a first course in noncommutative rings]. See in particular Theorem 7.19.


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.