in my research on dilations of contractions on Hilbert spaces and manifolds I have come across this nice publication concerning the classic Sz-Nagy theorem on the Arxiv by Levy and Shalit which states the classical Sz-Nagy theorem on the existence of unitary and isometric dilations for contractions and their minimality:


Right below quoting the famous theorem they mentioned that if the operator is not a unitary contraction then we must have that the larger Hilbert space K is infinite dimensional. Please forgive my ignorance but I cannot seem to understand why and it seems like a nice result I wish to understand. If the contraction to be dilated is not a unitary operator then why is the larger space embedding H is necessarily infinite dimensional? I am sorry if it was trivial and I missed it but I think I need to understand this. I thank all helpers


There are probably many ways to see this, but here is a quick one. Suppose $T$ is a contractive matrix which dilates to a unitary matrix $U$, we wish to show that $T$ is already unitary. Now, to say that the matrix $U$ dilates $T$ means there is an isometric matrix $V$ so that $$ T^n = V^* U^n V $$ for all $n\geq 0$. From this it follows that the minimal polynomial $p$ for $U$ must annihilate $T$, but since all the roots of $p$ are unimodular, it follows that all the eigenvalues of $T$ are unimodular, and thus $T$ is unitary.

  • $\begingroup$ Thank you Mike this is certainly a fresh approach for operator theory $\endgroup$ – kroner Mar 24 '16 at 14:41

A unitary dilation $U\in B(K)$ of $T\in B(H)$ is said to be minimal if $$\overline{span}\{U^nH : n\in \mathbb Z\} = K.$$ It is a well known result that the minimal unitary dilation is unique up to unitary equivalence.

For any contraction $T\in B(H)$, Schaeffer found a constructive version of Sz.-Nagy's dilation theorem. Specifically, $$ U = \left[\begin{array}{cccccc} \ddots \\ \ddots & 0\\ & 1& 0 \\ && D_{T^*} & T \\ && -T^* & D_T & 0 \\ &&&&1 & 0 \\ &&&&&\ddots&\ddots \end{array}\right] $$ is a unitary dilation of $T$, where $D_X = (1-X^*X)^{1/2}$.

Hence, let $K' = \overline{span}\{U^nH : n\in \mathbb Z\} \subset K$. $K'$ is a reducing subspace of $U$ and $U|_{K'}$ is the minimal unitary dilation of $T$. If $T$ is not a unitary already then either it isn't an isometry ($D_T \neq 0$), a co-isometry ($D_{T^*} \neq 0$) or both. Either way it is direct to see that $K'$ is infinite dimensional from the form of the Schaeffer dilation. Therefore, by minimality, any unitary dilation of $T$ must be on an infinite dimensional Hilbert space.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.