Linear independence of root equations forM/G/1 type Markov chains

There is a classical technique for determining the equilibrium probabilities of M/G/1 type Markov chains. After transforming the equilibrium balance equations of the chain, one obtains an equivalent system of equations in analytic functions to be solved. This method requires nding all singularities of a given matrix function in the unit disk and then using them to obtain a set of linear equations in the nite number of unknown boundary probabilities. The remaining probabilities and other measures of interest are then computed from the boundary probabilities. Under certain technical assumptions, the linear independence of the resulting equations is established by a direct argument involving only elementary results from matrix theory and complex analysis. Simple conditions for the ergodicity and nonergodicity of the chain are also given.