About the identification of $M_n(B(\mathcal{H}))$ with $B(\mathcal{H}^{(n)})$

81 Views Asked by At

I already checked another question where the *-isomorphism is shown.

As said in the book I'm reading now Completely bounded maps and operator algebras by Vern Paulsen page 3:

"There is a natural way to regard an element of $M_n(B(\mathcal{H}))$ as a linear map on $\mathcal{H}^{(n)}$, by using the ordinary rules for matrix products. That is, we set:

$$ (T_{ij}) \begin{pmatrix}h_1 \\ \vdots\\ h_n \end{pmatrix} = \begin{pmatrix}\sum_{j=1}^nT_{1j}h_j\\ \vdots\\ \sum_{j=1}^nT_{nj}h_j \end{pmatrix}$$

fot $(T_{ij})$ in $M_n(B(\mathcal{H}))$ and $(h_1, \dots h_n)^T \in \mathcal{H}^{(n)}$..".

What we should check now is that every element of $M_n(B(\mathcal{H}))$ defines a bounded linear operator on $\mathcal{H}^{(n)}$ obtaining the following inequality:

$$\|(T_{ij})\| \le \left(\sum_{i,j=1}^n\|T_{ij}\|^2\right)^{1/2}$$

Any help on how to do that?

EDIT:

That's what I'm thinking to do right now:

take $\bar{h}=(h_1, \dots, h_n) \in \mathcal{H}^{(n)}$ and consider

$$ \|(T_{ij})(\bar{h})\|^2 = \sum_{i=1}^n \left\|\sum_{j=1}^n T_{ij}h_j\right \|^2 \le \sum_{i,j = 1}^n \|T_{ij}h_j\|^2 \le \sum_{i,j =1}^n \|T_{ij}\|^2 \|h_j\|^2 $$

how to formally conclude?

Also the fact of considering elements of $\mathcal{H}^{(n)}$ as row vectors instead of columns vectors helped me quite. There would have been any difference?

Also, showing the isomorphism, I'd need some help for surjectivity : /

1

There are 1 best solutions below

0
On BEST ANSWER

Your first inequality is not true, the square of the norm does not satisfy the triangle inequality. What you can do is use the triangle inequality and then Cauchy=Schwarz as follows:
$$ \left\|\sum_{j=1}^n T_{ij}h_j\right \|^2 \leq\left(\sum_{j=1}^n \|T_{ij}\|\,\|h_j\|\right )^2\leq\sum_{j=1}^n\|T_{ij}\|^2\sum_{j=1}^n\|h_j\|^2. $$ Then $$\sum_{i=1}^n \left\|\sum_{j=1}^n T_{ij}h_j\right \|^2 \leq\sum_{i=1}^n\left(\sum_{j=1}^n\|T_{ij}\|^2\sum_{j=1}^n\|h_j\|^2\right) =\left(\sum_{i,j=1}^n\|T_{ij}\|^2\right)\sum_{j=1}^n\|h_j\|^2 $$

If your vectors are rows, the multiplication by a matrix is not "row times column", and it doesn't work on the left, the matrix $A$ acting on $x$ would be $xA^T$ (depending on how you define the matrix of an operator).

For the surjectivity, if $T\in B(H^n)$, if $\pi_j:H^n\to H$ is the projection onto the $j^{\rm th}$ coordinate, you can define $$T_{kj}h=\pi_k \circ T\begin{bmatrix} 0\\ \vdots \\ 0\\ h\\ 0\\ \vdots\\ 0\end{bmatrix},$$ where the $h$ is in the $j^{\rm th }$ coordinate.