2
$\begingroup$

If $n > 1$, there are no non-zero $*$-homomorphisms $M_n(\Bbb{C}) \to \Bbb{C}$. A $*$-homomorphism is an algebra morphism $\varphi: M_n(\Bbb{C}) \to \Bbb{C}$ with $\varphi(\overline{A}^T) = \overline{\varphi(A)}$

I tried to show that every $*$-morphism must be zero: if $\varphi$ is such a $*$-morphism, then $E_{ij}^2 = 0$ for $i \neq j$ (here $E_{ij}$ is the matrix that is $1$ on position $(ij)$, $0$ elsewhere). Thus $\varphi(E_{ij}) = 0$ for $i \neq j$. Hence, $$\varphi(A) = \sum_i a_{ii} \varphi(E_{ii})$$ so if I can show that $\varphi(E_{ii}) = 0$ I will be done. Unfortunately, I can't see why this should be true. I did not yet fully use the fact that it is a $*$-morphism. The only thing I can see is that $\varphi(A) \in \Bbb{R}$ for all $A$ since $\varphi(E_{ii})= \varphi(\overline{E_{ii}}^T) = \overline{\varphi(E_{ii}})$. If $\varphi$ is non-zero, then there is $A$ with $\varphi(A) \in \Bbb{R}\setminus \{0\}$. But then $\varphi(iA) = i\varphi(A) \notin \Bbb{R}$, a contradiction. Thus $\varphi=0$. Is this correct?

$\endgroup$
  • 2
    $\begingroup$ Yes, or just use the algebra homomorphism condition on $E_{ii}=E_{ij}E_{ji}$, $j\neq i$. $\endgroup$ – user10354138 Jul 29 at 8:13
  • $\begingroup$ @user10354138 So we actually have that every algebra morphism $M_n(\Bbb{C}) \to \Bbb{C}$ is zero? We don't need that it preserves the $*$-operations? $\endgroup$ – user745578 Jul 29 at 8:21
  • $\begingroup$ Why would you have $\phi(A)\in\mathbb{R}$? The diagonal entries are not necessarily real. The trace is a map that has all the properties you used, and is certainly not zero. $\endgroup$ – MaoWao Jul 29 at 8:39
  • $\begingroup$ @user1551 That is true, but I wrote "all the properties you used", and the OP did not use the multiplicativity of $\phi$ for anything but $\phi(E_{ij})=0$ for $i\neq j$, which is also true for the trace. $\endgroup$ – MaoWao Jul 29 at 8:43
  • 1
    $\begingroup$ @user1551 Once again, I know that the trace is not multiplicative. But the trace also has the property that $\phi(E_{ij})=0$ for $i\neq j$ and $\phi(E_{ii})\in \mathbb{R}$. These two facts are not sufficient to show that $\phi=0$, and that is what the OP tries to do. $\endgroup$ – MaoWao Jul 29 at 8:48
3
$\begingroup$

We can prove that when $n\ge2$ and $\mathbb F$ is a field, the only mapping $\phi:M_n(\mathbb F)\to\mathbb F$ that is additive and multiplicative is zero. (So, we don't need $\phi$ to be an algebra morphism, not to say a $\ast$-homomorphism over $\mathbb C$.)

For any $A\in M_n(\mathbb F)$, we have $A=P_1DQ$ for some nonsingular matrices $P_1,Q$ and some diagonal matrix $D$. Let $P_2$ be any permutation matrix with a zero diagonal. Then $P_2^TD$ has a zero diagonal too. It follows that $A=(P_1P_2)(P_2^TD)Q$ can be written as $A=P(L+U)Q$ where $P,Q$ are nonsingular, $L$ is strictly lower triangular and $U$ is strictly upper triangular.

Since $L^n=0$, we have $\phi(L)^n=\phi(L^n)=0$. Therefore $\phi(L)=0$ and similarly $\phi(U)=0$. Hence $\phi(A)=\phi(P)\left(\phi(L)+\phi(U)\right)\phi(Q)=0$.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ The above comment shows an easier way imo. Thanks though. $\endgroup$ – user745578 Jul 29 at 10:16

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.