A decomposition of the minimal Stinespring representation

117 Views Asked by At

The question is originated from an exercise of Paulsen's book "Completely Bounded maps and operator algebras" (P115 Exercises 8.2).

Let $A$ be a C*-algebra with unit, $H$ be a Hilbert space and let $\Phi: M_2(A)\rightarrow B(H\oplus H)$ be a completely positive map. Then we have a minimal Stinespring representation $(\pi_1, V, K_1)$ for $\Phi$. Prove the Hilbert space $K_1$ can be decomposed as $K_1=K\oplus K$ in such a way that the *- homomorphism $\pi_1:M_2(A)\rightarrow B(K\oplus K)$ has the form $$\pi_1\left[\left[ \begin{array}{ccc} a & b \\ c & d\\ \end{array}\right]\right]=\left[ \begin{array}{ccc} \pi(a) & \pi(b) \\ \pi(c) & \pi(d)\\ \end{array}\right],$$ where $\pi: A\rightarrow B(K)$ is a unital *-homomorphism.

My attempt: In this exercise, I do not know how to construct the Hilbert space $K$ such that $K_1=K\oplus K$. Because the Stinespring representation is minimal (i.e. $\overline{\pi_1(M_2(A))V(H\oplus H)}=K_1$), I guess $K=\overline{\pi_1(M_2(A))V(H\oplus{0})}\cong \overline{\pi_1(M_2(A))V(0\oplus H)}$. However, I am not sure whether it is true.

1

There are 1 best solutions below

4
On BEST ANSWER

Your approach sounds a priori like a good idea, but it cannot work because you have no control over $V$.

Instead, you have that $\pi_1(A)$ is non-degenerate by the minimality condition. This implies that $\pi_1(1)=1$ (because it is a projection with trivial kernel). Now let $$ K=\pi_1\left(\begin{bmatrix}1&0\\0&0\end{bmatrix}\right)\,K_1. $$ Then $$ K_1=K\oplus\pi_1\left(\begin{bmatrix}0&1\\1&0\end{bmatrix}\right)K\simeq K\oplus K.$$

Define

$$\pi(a)=\pi_1\left(\begin{bmatrix}a&0\\0&0\end{bmatrix}\right).$$ Then, writing $U=\pi_1\left(\begin{bmatrix}0&1\\1&0\end{bmatrix}\right)$, \begin{align} \pi_1\left(\begin{bmatrix}a&b\\c&d\end{bmatrix}\right) &=\pi(a)\pi(E_{11})+\pi(b)\pi(E_{12})+\pi(E_{21})\pi(c)+\pi(E_{21})\pi(d)\pi(E_{12})\\[0.3cm] &=\pi(a)+\pi(b)U+U\pi(c)+U\pi(d)U\\[0.3cm] &\simeq\begin{bmatrix}\pi(a)&\pi(b)\\\pi(c)&\pi(d)\end{bmatrix} \end{align}