The Setting
Rokhlin’s problem asks whether every mixing transformation is mixing of all orders. Let be an invertible measure-preserving transformation of a probability space , with integer powers , . A measure-preserving transformation of is mixing if (1) for all measurable sets and .
Higher orders are tested along layouts. Let be a finite set with . A layout sequence on is a sequence of maps , , and mixes along the proper sublayouts of if (2) for every proper subset of and all measurable sets , . An index is remote in if for every and every .
A witness of the failure of mixing of order along a layout sequence on is a family of functions bounded by , one of them of mean zero, whose correlations along , the integrals of the products over , stay above a constant . On a finite alphabet the functions are , , for one measurable observable with values in a finite set and functions , and the correlations are (3)
Coding and Word Entropy
The symbolic coding of by is the map given by (4)✦ The process of is the sequence of the coordinates of .✦
The entropy of a random variable with values in a finite set is the Shannon entropy of its law, (5)✦ with .
For a process with values in a finite set , indexed by or by , the word of on is the random variable with values in .✦ The word entropy of a process with values in a finite set is the entropy of its word of length from time , (6)✦ and has word-entropy rate zero if .
The Tail Factor
The completion of a sub--algebra of is the -algebra of sets in that agree -almost everywhere with a set in .✦ The past of a process with values in a finite set, at a time , is the -algebra generated by the words , .✦ The tail of is the -algebra (7)✦ The tail factor of the observable is the tail of the process of .✦
The clipped conditional expectation of a real function given a sub--algebra is its conditional expectation cut off at and , (8)✦
The Claim
Theorem 1. Let be a mixing invertible measure-preserving transformation of , let be a layout sequence on a finite set with such that mixes along the proper sublayouts of , let be a measurable observable with values in a finite set , and let .
If is remote in and for , then (9)✦
If is -measurable and for , then (10)✦
If is -measurable with values in a finite set , satisfies , and for , then (11)✦
If maps into a finite set , satisfies , the process has word-entropy rate zero, and for with , then the correlations (3) satisfy (12)✦ and there is no with (13)✦
In the paper, the result reads: if an invertible mixing transformation is not mixing of every order, then there are , a finite alphabet , a shift-invariant probability measure on and real functions of the coordinate such that the shift is mixing of every order below for , , , the entropy rate of is , and, for some and layouts whose least gap tends to infinity, for every , its being the page’s . The library states the result not as that existence statement but as its contrapositive: part 4 rules out, along layouts whose proper sublayouts mix, a finite-alphabet witness in which the function of mean zero is a function of a process of word-entropy rate zero.
What Is Checked
The theorem is stated over a mixing invertible measure-preserving transformation of a probability space, a layout sequence on at least three indices along whose proper sublayouts mixes, a remote index, a finite-valued observable with its tail factor , and the word entropy of a process coded by .
About these constructions the library states that the tail factor has zero entropy: for an invertible measure-preserving and a finite-valued observable , every -measurable with values in a finite set satisfies (14)✦ and, in integer time, the tail process satisfies (15)✦
For a decreasing sequence of sub--algebras of with , a real , indices and -measurable functions with , one has (16)✦
For an invertible measure-preserving , a layout sequence in which is remote, a finite-valued observable and functions , the factor at the remote index may be replaced in the limit by its clipped conditional expectation given : (17)✦