On quantum Markov chains on Cayley tree. II: Phase transitions for the associated chain with \(XY\)-model on the Cayley tree of order three (Q639267)

The paper under review concerns the phase transition for quantum Markov chains describing an \(XY\)-model on a regular tree of valence 4 (with one distinguished vertex with three edges). The model is described by an inductive limit \(C^{\ast}\)-algebra \(B\) arising as a tensor product of matrix algebras with a copy of \(M_2\) at every vertex in the tree (for analogous models see [\textit{O. Bratteli} and \textit{D. W. Robinson}, Operator algebras and quantum statistical mechanics. 2: Equilibrium states. Texts and Monographs in Physics. Berlin: Springer (1997; Zbl 0903.46066)]), with local Hamiltonians associated with an edge \(\langle u, v\rangle\) given by the expression \(H_{\langle u, v\rangle} = \frac{1}{2} \left(\sigma_x^{(u)} \sigma_x^{(v)} + \sigma_y^{(u)} \sigma_y^{(v)}\right)\), where \(\sigma_x, \sigma_y\) denote Pauli matrices, and the `transition' operators \(K_{\langle u, v\rangle} = \exp( \beta H_{\langle u, v\rangle})\) for a fixed \(\beta>0\). By a forward quantum Markov chain on such a model one understands a state on \(B\) which satisfies natural invariance properties for certain quasi-conditional expectations related to `cut-offs' of the tree at various levels. In earlier work the authors showed that such a state exists for any given family of `local' densities at each vertex satisfying explicit compatibility conditions. The appearance of the phase transition is understood here as the existence of two states on \(B\) which are forward quantum Markov chains for a given model and are not quasi-equivalent. The main results of the paper are the following: there exist constants \(0<\beta_*< \beta^*\) such that for \(\beta\in (\beta_*, \beta^*)\) we observe the phase transition, and for \(\beta>0\) outside that open interval the model admits a unique forward quantum Markov chain (at least if one assumes that the `local' densities depend only on the level of the corresponding vertex). This should be compared to the earlier work by the authors, where they showed that only the second type of behaviour is possible if we consider a tree of valence 3. The main part of the proof is based on a detailed analysis of a certain system of ordinary differential equations. For part I see [Infin. Dimens. Anal. Quantum Probab. Relat. Top. 14, No. 3, 443--463 (2011; Zbl 1239.46044)].