I am trying to understand different definitions of equivalence of representations. Given an algebra $\textbf{A}$, given two different representations $\pi_1$ and $\pi_2$ on two spaces $\textbf{B}(\textit{H}_1)$ and $\textbf{B}(\textit{H}_2)$, respectively. There are the following definitions:

  • $\pi_1$ and $\pi_2$ are said to be unitarily equivalent if there is a unitary opearator U such that U $\textbf{B}(\textit{H}_1)$ $U^{-1}$ = $\textbf{B}(\textit{H}_2)$. Otherwise, they are said to be not (unitarily) equivalent
  • $\pi_1$ and $\pi_2$ are said to be almost equivalent if every normal state for $\pi_1$ is also normal for $\pi_2$ and viceversa
  • $\pi_1$ and $\pi_2$ are said to be disjoint if there is no normal state for both representations

My doubts are the following:

  1. If both $\pi_1$ and $\pi_2$ are irreducible, $\pi_1$ and $\pi_2$ are unitarily equivalent $\iff$ $\pi_1$ and $\pi_2$ are almost equivalent?
  2. $\pi_1$ and $\pi_2$ are unitarily inequivalent $\iff$ $\pi_1$ and $\pi_2$ are disjoint?

Can you please help me? Even trivial examples would be helpful. My apologies in advance if I made stupid mistakes, I am still learning C* algebras.

Update: Theorem 2.4.26 from Bratteli-Robinson vol. I proves that almost equivalence $\iff$ existence of a isomorphism relating $\pi_1$ and $\pi_2$ as pointed out by @QuantumFieldMedalist. Moreover, it also proves that $\pi_1$ and $\pi_2$ are irreducible and almost equivalent $\iff$ unitarily equivalent


1 Answer 1


Unitary equivalence of representations is a stronger statement than almost equivalence. To see this, for two unitarily equivalent representations $\pi_{1}, \pi_{2}$, any normal state on one is unitarily related to a normal state on the other. Almost equivalence is weaker as the relation between the normal states is not necessarily implementable by a unitary.

Disjoint representations $\pi_{1}, \pi_{2}$ occur when the representations are not even quasi-equivalent. This condition means that there is no *-isomorphism between $\pi_{1}(A)'', \pi_{2}(A)''$. A unitary is only a special type of *-isomorphism, so two representations need not be disjoint if they are not unitarily equivalent.

  • $\begingroup$ Thank you! Just to better understand what you say: in case of almost equivalence, two normal states are related by a *-isomorphism which however is not unitary? May I ask you also a physical example? It would be helpful. $\endgroup$
    – MBlrd
    Commented Jun 6, 2023 at 19:46
  • $\begingroup$ The *-isomorphism is not necessarily unitary, and when it is then you have unitary equivalence instead. For a simple example, consider the algebra of observables of a QFT defined on a causal patch of de Sitter. This has an isomorphic algebra, namely the observables localized on the antipodal patch. These are isomorphic representations but are not related by a unitary, as one cannot evolve observables from one side to the other. $\endgroup$ Commented Jun 6, 2023 at 19:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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