Nonmonotonicity of phase transitions in a loss network with controls (Q862217)

The authors study a symmetric tree loss network that supports unicast and multicast calls to nearest neighbors and each link has the same capacity \(C\). The number of multicast calls centered at any node cannot exceed \(C_V\) and the number of unicast calls at a link cannot exceed \(C_E\) \((\max\{C_E,C_V\} \leq C)\). The first main result states that, given a pair of activity vectors (involved to construct a measure on an appropriate configuration space), the uniqueness of Gibbs measures for the unicast-multicast model is equivalent to the convergence of certain recursions of the explicitly specified map. In particular this assertion develops some results obtained by \textit{K. Ramanan, A. Sengupta, I. Ziedius} and \textit{P. Mitra} [Adv. Appl. Probab. 34, No. 1, 58--84 (2002; Zbl 1004.60051)] concerning the phase transitions occuring for the hard core model on the tree. The second principle result provides a precise characterization of the phase transition surface for the controlled unicast-multicast model under certain conditions.