1: \begin{abstract}

2: We study branching Markov chains on a countable state space $\Xs$, where the base Markov chain is

3: transient and irreducible. Our focus is on the limit behaviour of population averages, in

4: particular of the sequence of empirical distributions of the evolving population. The offspring

5: distributions at the points of $\Xs$ do not necessarily coincide, but are assumed to have

6: the same average and to be dominated by a probability on $\Zp$ which satisfies the $L \log L$

7: moment condition. In this case, we show that the associated population martingale is uniformly

8: integrable. Convergence of population averages is then connected with stationary spaces of

9: the base Markov chain, and in particular, with compactifications of $\Xs$ such that the base chain

10: converges almost surely to the boundary, resp. the Dirichlet problem at infinity is solvable.

11: Final considerations consider the general interplay between the measure theoretic boundaries of

12: the base chain and the branching chain. The general emphasis on qualitative rather than

13: quantitative aspects of the behaviour at infinity.

14: \end{abstract}