The Setting
A joining of two probability spaces defines a Markov operator between their spaces, sending a function of the second coordinate to its conditional expectation given the first. In Rokhlin’s multiple-mixing problem such operators come in families indexed by conditioning data, and four families connect four Hilbert spaces in a rectangle. The question is how each family acts on the essential ranges of the others and on their intersections at the corners, the spaces on which the rectangle is active.
The library treats an abstract rectangle of operator families over random parameters, and then a grouped square built from a random array, whose corners are spaces of blocks of cells and whose edges are joining operators of conditional laws of those blocks.
Operator Rectangles
Let be a measure on a measurable space , and let be a family of bounded linear operators between complex Hilbert spaces, indexed by the points of . The family is strongly measurable if, for every in , the map is strongly measurable.✦ The essential kernel of is the subspace of of the vectors that annihilates for -almost every ✦, and the essential range of is the closed subspace of orthogonal to the essential kernel of the adjoint family : (1)✦
The family has essential total range if the only closed subspace of that contains the range of for -almost every is itself: (2)✦
For a closed subspace of a Hilbert space, denotes the orthogonal projection onto . The compression of a bounded operator to closed subspaces and is the operator .✦
An operator rectangle is given by random elements , , , of a probability space , with laws , , , , such that each of the pairs , , , is independent, and by strongly measurable families of bounded operators , , and between complex Hilbert spaces, indexed by the values of the four random elements and satisfying both rectangle identities almost surely: (3)
Essential ranges of these families and of their adjoint families are taken for the law of the indexing random element, and so on. The active intersections are the closed subspaces (4) and the restricted operators , , and are the compressions of , , , to them.
Array Objects
Let be a measurable automorphism of a standard Borel space , a nonempty measurable space isomorphic to a Polish space with its Borel sets, and let be a finite index set. An array of order over a space is a function from the index tuples to , a point of .✦ A map acts on arrays entrywise✦, and an array of order zero is identified with its one entry, a point of .✦ For an axis and a slot , the plane restriction fixing axis at slot sends an array of order to the array of order whose entry at a tuple is the entry at that tuple with inserted at position ✦; the last-axis planes of an array along are its plane restrictions fixing the last axis at the slots in .✦
Initial array laws for are probability measures on , one for each order , that are invariant under acting entrywise, consistent, every plane restriction carrying to , and proper: for , the last-axis planes along are independent under , each with law .✦
An array object over the initial array laws for consists of standard Borel spaces of states and of parameters, each with an automorphism, measurable maps intertwining the automorphism of with and intertwining it with that of , and probability measures on , invariant, consistent and proper in the same sense and carried by entrywise to , such that and are independent under and the condition below holds.✦ The parameter array of an array is its image under entrywise, and is the conditional law under of an array of order given its parameter array.✦ The condition is that for the last-axis planes along of an array of order are independent given its parameter array, the plane at each having conditional law at the plane at of the parameter array.
An extension of by an array object over the same laws is a pair of measurable maps, one between the state spaces and one between the parameter spaces of and , that intertwine the automorphisms, are compatible with and , and whose state map carries the laws of entrywise to the laws of .✦ The pullback along the state map of an extension of by is composition with that map entrywise, a linear isometry of into .✦ On arrays of order an array object comes with sub--algebras of the Borel sets of : the parameter -algebra and, for each , the outside -algebra and the enlarged -algebra . The projection onto one of these -algebras is the conditional expectation onto it, the orthogonal projection of onto the functions measurable for it✦, and is saturated if these projections commute with pullback along every extension of .✦
The Grouped Square
A triple partition of is an ordered triple of nonempty, pairwise disjoint subsets of whose union is .✦ Fix an array object , a triple partition of serving as the rows, a triple partition of serving as the columns and a level , the order of the cells below. An array of order over has a cell for every row and column in , the array of order obtained by fixing its first axis at and its last axis at .✦ For sets of rows and of columns, is the block of the cells with and .✦ A group datum for and is a pair of a parameter array of order over and a block, a family of arrays of order over indexed by and ✦; the grouped datum of for and is the group datum formed by its parameter array and the block .✦
The row family kernel is the product kernel that sends a parameter array of order to the product, over the rows , of at the plane of fixing the first axis at .✦ For sets of rows and of columns, the group marginal kernel is the image of the row family kernel for under the map that keeps only the cells in the columns of ✦; its value at is the law of the cells of the block in the rows of and the columns of .✦
The joining operator of measure-preserving maps and from a probability space to and is the operator obtained by composing the pullback isometry with the adjoint of the pullback isometry along ; it sends to the conditional expectation of given .✦ A kernel that assigns to each point of a parameter space a coupling of two random points with laws and gives a family of joining operators , one for each , the joining operator of the coupling at , which sends a function of the second point to its conditional expectation given the first.✦
The row pair kernel for rows and columns , , is the conditional law under of the pair of blocks given the grouped datum of for and .✦ The row joining operator from the columns to the columns , conditioned on the columns , at a group datum , with a block in the rows of and the columns of , is the joining operator of the row pair kernel.✦ For fixed , the fibre row operators are the row joining operators at the group data , as a family over the blocks ✦; the row-active subspace of is their essential range for .✦
The parent rows of an array of order along a set are its plane restrictions fixing the first axis at the slots of .✦ The posterior kernel takes a parameter array of order and a family of arrays indexed by to the conditional law of an array drawn from given that its parent rows along are that family, the disintegration of the observed kernel, the joint law of the parent rows and the array.✦ For columns and rows , the column group kernel takes a group datum , with a block in the rows of and the columns of , to the product over of the posterior kernel at the plane of fixing the last axis at and the cells of in column .✦ For sets and of rows, the column pair kernel is the image of the column group kernel under the maps that keep only the cells in the rows and only those in the rows ; it is a coupling of the blocks and .✦ The column joining operator from the rows to the rows , conditioned on the rows , at is the joining operator of the column pair kernel.✦ For fixed , the column joining operators at the group data form the fibre column operators, a family over the blocks ✦, and the column-active subspace of is their essential range for .✦
Fix . At a corner of the square, with one of , and one of , , and with , the other ones, is the row-active subspace of the rows from the columns to conditioned on , and is the column-active subspace of the columns from the rows to conditioned on . Both are closed subspaces of , and their intersection is the common active space of the corner.✦ The edges of the square are the fibre row operators and of the rows and from the columns to conditioned on , and the fibre column operators and of the columns and from the rows to conditioned on , with , , , blocks in the rows and columns , , , . The restricted edge operators are the compressions of the edge operators to the common active spaces, , , and .
The Claim
Theorem 1. Let , , , and , , , form an operator rectangle. Then each family intertwines the projections onto essential ranges of the other families and of their adjoints, (5)✦ and the projections onto the active intersections (4), (6)✦ each identity holding for almost every value of the random element it involves.
Let be a saturated array object, and triple partitions of and a level. Then for almost every parameter array of order , under the law of the parameter array of an array drawn from , the row-active and column-active projections commute at the corner , (7)✦ the edge operators intertwine the projections onto the common active spaces, (8)✦ for almost every value of the block each identity involves, and the restricted edge operators satisfy both rectangle identities almost surely, (9)✦ each restricted family having essential total range (2) for the law of its block.
In the paper, the result reads: in every grouped cell, a corner of the square, the projections and commute; row operators intertwine the column-active projections and column operators the row-active ones; all incident operators intertwine the projections ; is invariant under conjugation; and the restrictions to the spaces have essential closed range equal to the target space, also for the reverse operators.
What Is Checked
The claim is stated over strongly measurable families of bounded operators between complex Hilbert spaces, indexed by random elements with four independent pairs and bound by the rectangle identities, with their essential kernels and ranges, active intersections and compressions; and over a saturated array object over initial array laws on a standard Borel space, two triple partitions of its index set, and the laws of cells, row and column joining operators and active subspaces built from its conditional laws.