Forbidden patterns and shift systems (Q2426426)

The authors study the occurence of order patterns in dynamical systems. Given an ordered set X , and a map T of the set to itself, then to every x element of X that is not periodic with period less then a certain L \(\geq 2\) an order pattern (permutation \(\pi_0,\dots,\pi_{L-1}\) of \(0,\dots,L-1)\) can be associated such that : \(T^{\pi_0} < T^{\pi_1} < \dots < T^{\pi_{L-1}}\). The authors show that under very relaxed conditions (a map that is piecewise monotone on a closed interval) not every permutation can occur (so there exist so-called forbidden patterns). Furthermore forbidden patterns induce larger forbidden patterns called the outgrowth. Root patterns are those forbidden patterns that belong not to the outgrowth of another forbidden pattern. In general the study of forbidden patterns tends to be very difficult. The authors therefor focus on the forbidden patterns of shift operators (one sided as well as two sided). For one sided shift operators they derive the following results: (A) One sided shift operators on \(N \geq 2\) symbols have no forbidden patterns of length \(L \leq N+1\). (B) Explicit construction of forbidden root patterns in case \(L\geq N+2\). These results immediately carry over to two sided shift operators by accomplishing a correspondence between patterns of one sided and two sided shift operators. The results can also be carried over to other dynamical systems using order isomorphisms. The paper is illuminated by lots of nice examples. I found the following two misprints: in the proof of Proposition 2 \(f_1^N(\phi(x_0))\) should be \(f_2^N(\phi(x_0))\), and in the middle of page 494 not \(\frac{5-\sqrt5}{8}\) is responsible for the splitting of \(P_{[1,0]}\) but \(\frac{5+\sqrt5}{8}\) is.