Approximating macroscopic observables in quantum spin systems with commuting matrices (Q1950180)

The global observables of an infinitely long spin chain are described by a \(C^*\)-inductive limit \(\mathcal A\) of \(\mathcal A_n = \bigotimes^n_{k=-n}B_k\), \(n\in {\mathbb N}\), with \(B_k = { M}_d(\mathbb{C})\), \(k\in \mathbb{Z}\). Let \(\gamma_j\in \mathrm{Aut}(\mathcal{A})\), \(j\in \mathbb{Z}\), map each \(B_k\) onto \(B_{k+j}\). Given \(m\in\mathbb{N}\) and s.a. \(A_i \in B_0\), \(1\leq i\leq m\), one has \([H_{n,i},H_{n,j}]\to 0\) as \(n\to \infty\), where for an \(m\)-tuple of averaged local observables \(H_{n,i} = (2n+1)^{-1}\sum_{j=-n}^n \gamma_j(A_i)\in \mathcal{A}_n\), \(n \in \mathbb{N}\). The main result is the existence of s.a. \(Y_{n,i}\in \mathcal{A}_n\), \(n\in \mathbb{N}\), \(1\leq i\leq m\), such that \(\|H_{n,i}-Y_{n,i}\|\to 0\), \(n\to \infty\), and \([Y_{n,i},Y_{n,j}]=0\) for all \(n,i,j\). The proof involves fine entropy estimates of projectors ranks. Notably, due to H.-X. Lin, an \(m\)-tuple of asymptotically commuting matrix sequences can be norm-approximated by a commutative \(m\)-tuple if \(m=2\). However, for \(m>2\), this fails in general.