We prove existence of asymptotic entropy of random walks on regular languages over a finite alphabet and we give formulas for it. Furthermore, we show that the entropy varies real-analytically in terms of probability measures of constant support, which describe the random walk. This setting applies, in particular, to random walks on virtually free groups.the concatenated word. A random walk on a regular language is a Markov chain (X n ) n∈N 0 on the set A * = n≥1 A n ∪ {o}, whose transition probabilities obey the following rules:(i) Only the last two letters of the current word may be modified.(ii) Only one letter may be adjoined or deleted at one instant of time. (iii) Adjunction and deletion may only be done at the end of the current word. (iv) Probabilities of modification, adjunction or deletion depend only on the last two letters of the current word and on the substitute letters.
Compare with Lalley [14]and Gilch [6]. In other words, at each step the last two letters of the current word may be replaced by a non-empty word of length of at most 3 and the transition probabilities depend only on the last two letters of the current word and the replacing word of length of at most 3. More formally, the transition probabilities of the Markov chain (X n ) n∈N 0 can be written as follows, where w ∈ A * , a 1 , a 2 , b 1 , b 2 , b 3 ∈ A:Not all of these probabilities need to be strictly positive. Initially, we set X 0 := o. If we start the random walk at w ∈ A * instead of o, we write P w [ · ] := P[ · | X 0 = w]. For w 1 , w 2 ∈ A * , the n-step transition probabilities are denoted by p (n) (w 1 , w 2 ) := P w 1 [X n = w 2 ]. The set of accessible words from o is given byWe will also think of the random walk (X n ) n∈N 0 as a nearest neighbour random walk on an undirected graph G, where the vertices are the elements of L and undirected edges are between two vertices if and only if one can walk from one word to the other one in a single step. For this purpose, we need the following assumption:We call this property weak symmetry.In particular, Assumption 2.1 yields irreducibility of the random walk on L. Moreover, this assumption will be necessary for the construction of a sequence of cones in the graph G which track the random walk's way to infinity. As the interested reader will see, weak symmetry can obviously be weakened in some way but for reason of better readability we keep this natural assumption; for a discussion on this assumption, we refer to Appendix A.2.Since the purpose of the paper is the investigation of the asymptotic behaviour of transient random walks, we obviously need that L is infinite in our setting. It is an easy exercise