Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

I have been reading the paper - "Introduction to Quantum Fisher Information". In section 1.2 the author talks about the linear map $\mathbb{J}_D$, which he defines as follows:

Let $D \in M_n$ be a positive invertible matrix. The linear mapping $\mathbb{J}^f_D:M_n \to M_n$ is defined by the formula

$\mathbb{J}_D^f=f(\mathbb{L}_D\mathbb{R}^{-1}_D)\mathbb{R}_D$

where $f:\mathbb{R}^+\to\mathbb{R}^+$,

$\mathbb{L}_D(X)=DX$ and $\mathbb{R}_D(X)=XD$.

Also, the author points out that the inverse mapping of $\mathbb{J}_D^f$ is the given by

$\frac{1}{f}(\mathbb{L}_D\mathbb{R}^{-1}_D)\mathbb{R}_D^{-1}$

Then he gives the following example, of which I am able to get the first part but not the inverse map.

Example 1: If $f(x)=(x+1)/2$, then

$\mathbb{J}_DB=\frac{1}{2}(DB+BD)$.

$\mathbb{J}_DB=\int_0^\infty \text{exp}(-tD/2)B\text{exp}(-tD/2)dt$

Can you tell me how does he get the required formula for the inverse map? Also, I would like to know what is the meaning of the map $\mathbb{J}_D$? What does the integral of the linear map represent? Can you also provide me some reference book where I could look for these topics?

share|improve this question

1 Answer 1

up vote 0 down vote accepted

One way to think of the operation ${J}_D^f$ is as follows: $M_n$ has the structure of a Hilbert space by the inner product $\langle X, Y \rangle = {\rm Tr}(X^* Y) = {\rm Tr}(Y X^*)$. Then the transformations $L_D R_{D}^{-1}$ and ${R}_D$ are positive operators on this Hilbert space. Since they commute, you may take the functional calculus $f(s) \otimes g(t) \mapsto f(L_D R_{D}^{-1}) g({R}_D)$ from the space of functions on ${\rm Sp}(L_D R_D^{-1}) \times {\rm Sp}(R_D)$ into the space of bounded operators on $M_n$ (since the spectrum ${\rm Sp}(L_D{R}_{D^{-1}})$ of $L_D$ is a finite subset of ${R}$, there is no need to worry about the regularity of $f$. Ditto for $g$). Then ${J}_D^f$ is simply the image of $f(s) \otimes t$. Since the functional calculus is an algebra homomorphism, the inverse of ${J}_D^f$ is represented by $(1/f(s)) \otimes (1/t)$, which equals $\frac{1}{f}(L_D{R}_D^{-1}){R}_D^{-1}$. When $E \subset M_n$ is the joint eigenspace for $L_D {R}_{D^{-1}}$ and ${R}_D$ with the corresponding eigenvalues $a b^{-1}$ and $b$ (i.e. $L_D$ has eigenvalue $a$ on this subspace), ${J}_D^f$ acts by $f(a/b)b$ on $E$.

Any book containing the spectral theory of self adjoint operators on Hilbert spaces will do, like Pedersen's Analysis Now (GTM 118).

About the computation of $\mathbb{J}_D$ for the case of $f(t) = \frac{t+1}{2}$, the integral formula follows from the identity $\int_0^\infty exp(-t a / 2) exp(-t b / 2) = \frac{2}{a+b} = \frac{2}{ab^{-1}+1} \frac{1}{b}$.

Edit: I made a stupid mistake in the first version, and this is the corrected version. Sorry for the change of notation from $\mathbb{L}_D$ to $L_D$, etc. I somehow couldn't make it to work.

share|improve this answer
    
Thanks for your reply. But I am still not clear on the computation. After applying the integral formula I have the following expression $\left(\int_0^\infty exp(-t\mathbb{L}_D/2)exp(-t\mathbb{R}_D/2) dt\right)(B)$. How does one evaluate this? –  Shishir Pandey Oct 26 '10 at 10:48
    
If $B$ satisfies $\mathbb{L}_D(B) = a B$ and $\mathbb{R}_D(B) = b B$, you have $exp(-t\mathbb{L}_D/2)exp(-t\mathbb{R}_D/2)(B) = exp(-ta/2)exp(-tb/2) B$ for any $t$. Its integral is, as above, $\frac{1}{f(ab^{-1})} b^{-1} B$, which equals $\frac{1}{f}(\mathbb{L}_D\mathbb{R}_D^{-1})\mathbb{R}_D^{-1}(B) = \mathbb{J}_D(B)$. For general matrix $B$, you can decompose it as a sum of the simultaneous eigenvectors for $\mathbb{L}_D$ and $\mothbb{R}_D$ because they are commuting positive operators. Functional calculus is compatible with this decomposition. –  Makoto Yamashita Oct 27 '10 at 1:50

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.