An entropy estimate for infinite interval exchange transformations (Q455620)

In this paper the author investigates the minimal entropy convergence rates for infinite interval exchange transformations (i.e.t. for short) and give some sufficient condition for the entropy of the i.e.t. to vanish. Suppose the interval \(I=[0,1]\) is partitioned into subintervals \(I=\bigsqcup_{k\geq1}I_k\) (indexed in descending order according to the Lebesgue measures of \(I_k\)). Define a map \(T:I\to I\), \(x\mapsto x+d_k\) (where \(x\in I_k\) and for some real numbers \(d_k\)) in such a way that \(T\) is injective. Clearly \(T\) preserves the Lebesgue measure (fix this measure from now on). Let \(p(x)=|I_k|\) if \(x\in [k,k+1)\), and \(q(x)\) be the unique number fulfilling the equation \(\displaystyle \int_{q(x)}^\infty p(d)dt=\frac{1}{x}\). Then the main result in this paper is: { Theorem 1.3.} Let \(\alpha\) be a partition of \(I\) into finitely many subintervals. Then \[ \limsup_{n\to\infty}\frac{H(\alpha_0^{n-1})}{2\log_2(q(n))+4\log_2(n)}\leq 1. \] Then the author gives several applications to the entropy of \(T\). It is easy to see that the entropy is zero for all finite i.e.t., which may not be the case of infinite i.e.t.. A direct corollary of Theorem 1.3 is: { Theorem 3.1.} If the sequence \(q(n)\) satisfies \(\limsup_{n\to\infty}\frac{\log q(n)}{n}=0\), then \(h(T)=0\). Let \(f(x)=-x\log_2(x)\). The author also gives a sufficient condition for vanishing entropy: { Corollary 3.3.} If \(\sum_{k\geq1}f(|I_k|)<\infty\), then \(h(T)=0\). Intuitively, it says that, finite information of the generating partition implies zero entropy of the system.