Rank One Transformations

Updated 16 January 2026

Rank one transformations are invertible measure-preserving systems built via iterative cutting-and-stacking that form towers approximating the entire measurable space.
They exhibit key ergodic properties such as minimal self-joinings, trivial centralizers, and simple spectral types that serve as benchmarks in dynamical systems.
These transformations are instrumental in analyzing mixing, rigidity, and decomposition in both finite and infinite measure settings, with applications spanning symbolic and topological models.

A rank one transformation is an invertible measure-preserving transformation on a standard (typically nonatomic) Lebesgue space constructed via a cutting-and-stacking procedure yielding, at each stage, a single tower whose partitions increasingly approximate the full measurable space. This class encapsulates some of the simplest yet structurally rich models in ergodic theory, providing key examples and counterexamples for problems concerning mixing, spectral properties, joinings, and centralizers.

1. Construction and Definition

A rank one transformation $T: X \to X$ (with $(X, \mathcal B, \mu)$ a standard Lebesgue space—finite or $\sigma$ -finite infinite) is constructed inductively:

Cutting-and-stacking scheme: Start with a base interval $B_0$ (finite measure, often normalized to $1$), forming a column $C_0 = \{B_0, T(B_0), \ldots, T^{h_0-1}(B_0)\}$ of height $h_0=1$ .
At stage $n$ , cut the base $B_n$ of column $C_n$ into $r_n \geq 2$ equal subintervals $B_{n,0}, \ldots, B_{n, r_n-1}$ .
Spacers: For each $0 \leq i < r_n$ , append $s_n(i)\geq 0$ "spacer" intervals of the same width above the $i$ th subcolumn. Stack the subcolumns (with spacers) left to right, obtaining a new column $C_{n+1}$ of height $h_{n+1} = r_n h_n + \sum_{i=0}^{r_n-1} s_n(i)$ .
The transformation $T$ is defined as the pointwise limit, acting as the level-to-level shift within each tall column and iteratively extending as the union of levels covers $X$ .

Variants include modifications for infinite measure, symbolic codings via rank-one words, and explicit (C,F)-models for group actions (Gaebler et al., 2016, Danilenko, 2016, Ryzhikov, 2020).

2. Key Structural and Ergodic Properties

Rank one transformations serve as canonical examples for core dynamical concepts:

Minimal self-joinings (MSJ): For many bounded, non-rigid, totally ergodic rank-one transformations, the only ergodic self-joinings are the diagonal and product measures. Such systems have trivial centralizer, are prime, and have strong independence properties among their powers (Gao et al., 2013, Ryzhikov, 2020).
Centralizer Triviality: Trivial centralizer ( $C(T) = \{T^n\}_{n\in\mathbb Z}$ ) holds under conditions such as bounded cuts or (in the infinite-measure case) the partial boundedness property. The latter demands uniform boundedness on cuts and bounded difference between spacers, together with a lower bound on spacer size in terms of the tower height, for all sufficiently large $n$ (Gaebler et al., 2016).
Genericity: The set of rank one transformations is a dense $G_\delta$ in the space of all measure-preserving transformations, and generic mixing transformations are rank one (Bashtanov, 2012, Gao et al., 2013, Neretin, 2018). Generic properties include ergodicity, weak mixing, trivial centralizer, and MSJ.

3. Spectral Theory and Rigidity

Spectral Type: For a generic (in Baire category) rank one transformation, the Koopman operator $U_T$ has simple, purely singular maximal spectral type, given explicitly by a Riesz product derived from the tower construction data (Neretin, 2018). The spectral multiplicity is typically one, and all convolution powers of the maximal spectral measure are pairwise singular.
Spectral Disjointness: For a broad class of rank one transformations (bounded-recurrent, non-flat parameters), the continuous spectra of different positive powers $T^p, T^q$ are mutually singular. The analysis utilizes the existence of analytic functions in the weak closure of $\{U_T^n\}$ and a combinatorial approach to the weak limits of powers (Abdalaoui et al., 2013).
Rigidity and Non-Rigidity: Rank one maps may be rigid, non-rigid, or partially rigid depending on spacer and cut growth. Infinite-measure, non-rigid constructions can yield totally ergodic but non-weakly mixing actions with nontrivial Krieger type (including type III) (Drillick et al., 2018). Full rigidity implies (via King, Ryzhikov) all self-joinings arise as strong operator limits of convex combinations of powers of $U_T$ (Chaika, 2019).

4. Classification, Isomorphisms, and Factors

Rank one transformations exist in both finite and infinite measure, with factor/isomorphism structure determined by explicit combinatorial criteria:

Finite cyclic and odometer factors: A rank one $T$ factors onto a finite cyclic permutation or odometer system if, asymptotically over tower levels, indices concentrate mod $k$ (cyclic), or for all $k$ in some infinite divisor-closed set (odometer) (Foreman et al., 2019).
Isomorphism to inverse: For partially bounded infinite-measure systems, $T\cong T^{-1}$ if and only if the spacer vector is eventually palindromic for large $n$ (Gaebler et al., 2016).
Isomorphism and disjointness criteria: For bounded-parameter, nonrigid, totally ergodic rank-one transformations, isomorphism reduces to eventual agreement of spacer sequences, and disjointness to eventual disparity (Danilenko, 2016).

5. Dynamical and Statistical Implications

Rational ergodicity: All rank-one transformations are subsequence boundedly rationally ergodic, and with bounded cutting numbers, satisfy bounded rational ergodicity (Bozgan et al., 2013). However, not all are weakly rationally ergodic; the weak RE property is meager among rank-ones.
Mixing: Generic mixing is rank one in Tikhonov's leash-topology; conjugacy classes of all mixing transformations are dense in this topology (Bashtanov, 2012).
Minimal self-joinings and mixing orders: In the bounded, nonrigid, mixing case, all symmetric powers are simple and MSJ of all orders holds (Ryzhikov, 2020, Gao et al., 2013).

6. Symbolic and Topological Models

Symbolic codings: Every cutting-and-stacking procedure yields a symbolic rank-one word, providing the basis for an explicit model space of symbolic rank-one measure-preserving systems with a natural Polish topology. These models mirror all generic properties of geometric rank-one transformations (Gao et al., 2013).
Nearly continuous Kakutani equivalence: All strongly rank-one systems (including the classical Chacon map) are nearly continuously even Kakutani equivalent to irrational rotations, unifying a wide class of symbolic and geometric models in ergodic theory (Rudolph et al., 2012).

7. Applications, Variants, and Open Problems

Matrix and tensor decompositions: In linear algebra and random matrix theory, the concept of rank-one transformations appears in the context of rank-one updates and decompositions. For example, explicit SVD-like decompositions for tensor operators and the preservation of polynomial ensemble structure under rank-one modifications (Kuijlaars, 2015, Turchetti, 2023).
Interval exchange transformations: Almost every irreducible interval exchange transformation is totally rank one—all powers are rank one and these are dense in the weak closure (Wu et al., 2016).
Open questions: Problems remain in the explicit description of weak closures, extension to higher-rank constructions with strong self-joining and spectral properties, existence of infinite-measure rank-one systems with Lebesgue spectrum, and the full structure of homoclinic groups and weak limits of powers in various settings (Ryzhikov, 2020).

References: