[Paper Review] On Mixing Constructions with Algebraic Spacers

V. V. Ryzhikov|arXiv (Cornell University)|Aug 6, 2011

Advanced Topology and Set Theory2 references3 citations

TL;DR

This paper establishes mixing in rank-one transformations by replacing stochastic spacers with algebraic spacers derived from multiplicative group generators in finite fields. Using the weak mixing property and the equidistribution of algebraic spacer differences, it proves that such constructions are mixing when the spacer sequences satisfy specific injectivity and boundedness conditions, extending Ornstein's stochastic mixing result to a deterministic algebraic setting with explicit eigenvalue control.

ABSTRACT

We recall a proof of the mixing for almost all Ornstein's stochastic rank one constructions, replace stochastic spacers by special algebraic ones and prove the mixing in this new situation.

Motivation & Objective

• To extend Ornstein's stochastic mixing result for rank-one constructions to a deterministic algebraic setting.
• To construct rank-one transformations using algebraic spacers derived from primitive roots modulo primes.
• To prove that such algebraic constructions are mixing under the condition of weak mixing and specific structural properties of spacer sequences.
• To eliminate the possibility of nontrivial eigenvalues—particularly −1—by controlling the parity of spacer differences.
• To provide a constructive, number-theoretically grounded alternative to probabilistic spacer selection in mixing rank-one systems.

Proposed method

• Define spacer sequences using residues of powers of a primitive root $ q_j $ modulo prime $ r_j $, setting $ s_j(i) = r_j + \{q_j^i\} - \{q_j^{i+1}\} $.
• Establish two key properties: (1) bounded partial sums $ |S_j(i,n)| \leq r_j $, and (2) injectivity of $ S_j(i,n) $ over $ i $, ensuring uniform distribution.
• Use weak mixing of the transformation to rule out nontrivial eigenfunctions, particularly focusing on the $-1$ eigenvalue.
• Apply a spectral argument: if $ Q(j) = \frac{1}{r(j)}\sum_{i=1}^{r(j)} T^{S_j(i,n)} \not\to \Theta $, then $ Q $ defines a $ T \otimes T $-invariant measure $ \eta \ll \mu \times \mu $, contradicting weak mixing unless $ Q = \Theta $.
• Leverage Burgess's bound on primitive roots to ensure $ q_j < \sqrt{r_j} $, enabling control over the parity distribution of $ \{q_j^i\} - \{q_j^{i+1}\} $.
• Show that the imbalance in parity counts $ ||M_0| - |M_1|| $ is sublinear, preventing $ (-1)^{S(i,1)} $ from being constant over large sets, thus eliminating $-1$ as an eigenvalue.

Experimental results

Research questions

• RQ1 Can mixing in rank-one systems be established using deterministic algebraic spacers instead of stochastic ones?
• RQ2 What structural conditions on algebraic spacer sequences ensure mixing in rank-one constructions?
• RQ3 How can the weak mixing property be preserved when replacing random spacers with algebraic ones?
• RQ4 Can the eigenvalue $-1$ be eliminated in such algebraic constructions through number-theoretic control of spacer differences?
• RQ5 What role does the equidistribution of $ S_j(i,n) $ play in the spectral convergence of averaging operators to the projection $ \Theta $?

Key findings

• The construction using algebraic spacers $ s_j(i) = r_j + \{q_j^i\} - \{q_j^{i+1}\} $ satisfies the boundedness and injectivity conditions $ |S_j(i,n)| \leq r_j $ and $ S_j(i,n) \neq S_j(m,n) $ for $ i \neq m $, which are essential for spectral approximation.
• The averaging operator $ Q(j) = \frac{1}{r(j)}\sum_{i=1}^{r(j)} T^{S_j(i,n)} $ converges weakly to $ \Theta $, the projection onto constants, under the assumption of weak mixing.
• The weak mixing of the system implies that any weak limit of $ Q(j) $ must be $ \Theta $, as otherwise a non-trivial $ T \otimes T $-invariant measure $ \eta \ll \mu \times \mu $ would contradict ergodicity of $ T \otimes T $.
• The parity of $ \{q_j^i\} - \{q_j^{i+1}\} $ is controlled via interval partitioning: for $ q_j < \sqrt{r_j} $, the imbalance in even and odd differences is at most $ \sqrt{r_j} $, preventing $ (-1)^{S(i,1)} $ from being constant.
• The construction avoids the $-1$ eigenvalue because the set of $ i $ for which $ S(i,1) $ is even or odd is asymptotically balanced, ensuring no eigenfunction exists.
• The result establishes that algebraic spacers can replace stochastic ones in proving mixing, with the key advantage of explicit, number-theoretically grounded control over spacer behavior.