Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

For $A,B,L$ linear operators, when is there a linear operator $X\{A,B\}$ such that

$$ALA^{-1}+BLB^{-1}=2 XLX^{-1}$$

can be solved independently for all $L$ only depending on $A$ and $B$?

share|cite|improve this question
Looks awfully like a generalized Sylvester equation... – J. M. Aug 9 '11 at 8:08
If the equation $(a_{ij}b_{kl}-a_{kl}b_{ij})^2=0$ holds for all $i,j,k,l$, and the matrix entries are real or complex numbers, does it not follow that $B$ and $A$ linearly dependent? – Jyrki Lahtonen Aug 9 '11 at 8:20
@JyrkiLahtonen: please disregard my 2x2 case, that was calculated when I omitted the $\frac12$ and assumed $L\neq1$. But maybe the answer still boils down to linear dependence – Tobias Kienzler Aug 9 '11 at 8:36

For any fixed $k \neq 2$, there does not exist any $X \in GL(n,\mathbb{C})$ such that $$ A L A^{-1} + B L B^{-1} = k X L X^{-1}$$ for all $L \in M_n(\mathbb{C})$, for if there were, then taking $L=1_n$ yields $$ 0 = A 1_n A^{-1} + B 1_n B^{-1} - k X 1_n X^{-1} = (2-k) 1_n,$$ which is impossible. In particular (unless I've misunderstood something), no such $X$ exists in your case, which is $k = \tfrac{1}{2} \neq 2$.

EDIT: In the case that $k = 2$, since every algebra automorphism of $M_n(\mathbb{C})$ is inner, and since $ALA^{-1} + BLB^{-1} = A(L + CLC^{-1})A^{-1}$ for $C = A^{-1}B$, your problem is equivalent to finding all $C \in GL(n,F)$ such that $L \mapsto \tfrac{1}{2}(L + CLC^{-1})$ is an algebra automorphism of $M_n(\mathbb{C})$. In the case that $C^2=1_n$, then you can check that this is true if and only if $C$ is a non-zero scalar multiple of the identity, in which case $\tfrac{1}{2}(L+CLC^{-1}) = L$, so that $B$ must be a non-zero scalar multiple of $A$, and you can just take $X=A$. Anything more general looks like a bit of a mess, though perhaps you can compute something explicit for $n=2$?

share|cite|improve this answer
The $k^{-1}$ in the displayed equation should ber $k$, no? – coffeemath Jan 1 '13 at 17:11
Indeed. Thanks for the correction! – Branimir Ćaćić Jan 1 '13 at 17:20
+1 Urgh, I must have slept when posting this question... Thanks for your input, please apologize that in that case I change the question into the one with correct factor... – Tobias Kienzler Jan 1 '13 at 17:21
I was actually wondering why you wanted a $\tfrac{1}{2}$ there, but with a $2$ everything makes sense. I've been trying to apply the necessary condition that $L \mapsto \tfrac{1}{2}(ALA^{-1}+BLB^{-1})$ be multiplicative, but all that seems to imply is something like $ALMA^{-1} + BLMB^{-1} = ALA^{-1}BMB^{-1} + BLB^{-1}AMA^{-1}$; plugging $L = 1$ or $M = 1$ yields nothing new, whilst plugging in, say, $L = A^{-1}$ and $M = A$ doesn't seem to yield anything helpful. – Branimir Ćaćić Jan 1 '13 at 17:30

Your Answer


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.