a b s t r a c tIn a recent paper with L.Q. Zamboni, the authors introduced the class of ϑ-episturmian words. An infinite word over A is standard ϑ-episturmian, where ϑ is an involutory antimorphism of A * , if its set of factors is closed under ϑ and its left special factors are prefixes. When ϑ is the reversal operator, one obtains the usual standard episturmian words. In this paper, we introduce and study ϑ-characteristic morphisms, that is, morphisms which map standard episturmian words into standard ϑ-episturmian words. They are a natural extension of standard episturmian morphisms. The main result of the paper is a characterization of these morphisms when they are injective. In order to prove this result, we also introduce and study a class of biprefix codes which are overlap-free, i.e., any two code words do not overlap properly, and normal, i.e., no proper suffix (prefix) of any code-word is left (right) special in the code. A further result is that any standard ϑ-episturmian word is a morphic image, by an injective ϑ-characteristic morphism, of a standard episturmian word.
Proposition 4.2. Let Z be a suffix, left normal, and overlap-free code over A, and let a, b ∈Then we can assume that (20) holds for suitable h ≥ 0, δ ∈ A * , and ξ ∈ A + . We have n > 1, for otherwise the statement is trivial, and δ = ε since va / ∈ Z * . As δ = a, if |δλ| ≤ |z| we obtain aλ ∈ Fact Z and we are done. Therefore assume |δλ| > |z|. In this case ξ is a proper prefix of λ and a proper suffix of z. If λ ∈ Pref Z * we reach a contradiction, since ξ ∈ Suff Z ∩ Pref Z * and this contradicts the hypothesis that Z is a suffix and overlap-free code. Thus λ / ∈ Pref Z * ; this implies, by the previous lemma, that aλ ∈ Fact Z . Proposition 4.3. Let Z be a biprefix, overlap-free, and right normal code over A. If λ ∈ Pref Z * \ {ε}, then there exists a unique word u = z 1 · · · z k with k ≥ 1 and z i ∈ Z , i = 1, . . . , k, such that u = z 1 · · · z k = λζ , z 1 · · · z k−1 δ = λ,where δ ∈ A + and ζ ∈ A * .Proof. Let us suppose that there exist h ≥ 1 and words z 1 , . . . , z h ∈ Z such that z 1 · · · z h = λζ , z 1 · · · z h−1 δ = λ (22) with ζ ∈ A * and δ ∈ A + . From (21) and (22) one obtains u = z 1 · · · z k = z 1 · · · z h−1 δ ζ and z 1 · · · z h = z 1 · · · z k−1 δζ , with z k = δζ and z h = δ ζ . Since Z is a biprefix code, we derive h = k and consequently z i = z i for i = 1, . . . , k − 1. Indeed, if h = k, we would derive by cancellation that δ ζ = ε or δζ = ε, which is absurd as δ, δ ∈ A + .Hence we obtain z k = δ ζ = δζ , whence δ = δ . Thus δ is a common nonempty prefix of z k and z k . Since Z is right normal, by Proposition 1.1 we obtain that z k is a prefix of z k and vice versa, i.e., z k = z k .Proposition 4.4. Let Z be a biprefix, overlap-free, and normal code over A. If u ∈ Z * \ {ε} is a proper factor of z ∈ Z , then there exist p, q ∈ Z * , h, h ∈ A + such that h / ∈ Suff Z , (h ) f / ∈ Pref Z , and z = hpuqh .Proof. Since u is a proper factor of z ∈ Z , there exist ξ , ξ ∈ A * such that z = ξ uξ ; moreover, ξ...