Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

If $V_0$ is the subspace consisting of matrices of the form $C=AB-BA$ for some $A,B$ in a vector space $V$ then $V_0=\{A\in V|\operatorname{Trace}(A)=0\}$.

The problem above is one of the past qualifying exam problems. I can prove that $V_0\subset\{A\in V|\operatorname{Trace}(A)=0\}$ but I do not know what to do with converse.

Thank you in advance.

share|improve this question
1  
What have you tried? –  Qiaochu Yuan Aug 11 '12 at 20:44
8  
14% accept rate? That'd make any matrix go singular... –  DonAntonio Aug 11 '12 at 20:45
    
Hint: Dimension count. –  Ehsan M. Kermani Aug 11 '12 at 21:24

2 Answers 2

up vote 7 down vote accepted

Here is a geometric approach:

Let $\langle A, B \rangle = \mathbb{tr} (A^* B)$. It is straightforward to check that this is an inner product on the set of $p \times p$ matrices.

Let $V_1 = \{A\in V|\operatorname{Trace}(A)=0\}$. It is clear that $V_0 \subset V_1$; we need to show the reverse inclusion.

Note that $V_1 = \{ A | \langle I, A \rangle = 0 \}$, where $I$ is the identity. Also note that $V_0, V_1$ are both (closed) subspaces. To show inclusion it suffices to show that $V_0^\bot \subset V_1^\bot$. It is easy to see that $V_1^\bot = \mathbb{sp} \{I \}$.

Suppose $X \in V_0^\bot$. Then, for any matrices $A,B$, we have $\langle X, AB \rangle = \langle X, BA \rangle$. Let $E_{ij} = e_i e_j^T$, and choose $A= E_{ij}, B = E_{mn}$. Then we get $\langle X, AB \rangle = \langle X, E_{ij} E_{mn} \rangle = \delta_{jm} [X]_{in}$. Similarly, we get $\langle X, BA \rangle = \delta_{in} [X]_{mj}$. Consequently we have $\delta_{jm} [X]_{in} = \delta_{in} [X]_{mj}$ for all $i,j,m,n$. Now choose $j=m$, this simplifies to $ [X]_{in} = \delta_{in} [X]_{jj}$. If $i \neq n$, this gives $[X]_{in} = 0$, if $i = n$, this shows that all diagonal elements are equal, so we have $X = \lambda I$, for some $\lambda$. Consequently $X \in V_1^\bot$.

share|improve this answer
    
+1, nice solution! –  joriki Aug 12 '12 at 10:03

Take $B$ to be diagonal with distinct diagonal entries $b_1,\ldots, b_n$. Then the $ij$th entry of $AB-BA$ is $a_{ij}(b_i-b_j)$. Thus, with this particular $B$, the map $A\rightarrow AB-BA$ is injective on matrices $A$ with diagonal entries $0$. (And dimension count.)

Edit: and as in @copper.hat's comment/question, this does not instantly address the diagonal matrices, and/but that reduces to the 2-by-2 case. Indeed, perhaps this is the more interesting part of the question. But/and the main point is that, with $x=\pmatrix{0&1\cr 0 & 0}$ and $y=\pmatrix{0&0\cr 1&0}$, the commutator is $xy-yx=\pmatrix{1&0\cr 0&-1}$.

Extrapolating this gives the general diagonal things.

share|improve this answer
1  
How does this show, for example, that $\mathbb{diag}(1,-1,0,...) \in V_0$? –  copper.hat Aug 12 '12 at 0:09

Your Answer

 
discard

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.