- The paper establishes the existence of unique absolutely continuous invariant probability measures for greedy and lazy double-base expansions using Perron-Frobenius theory and the Lasota-Yorke framework.
- It shows that both the greedy and lazy maps are exact, mixing, and ergodic, confirming strong stochastic behavior in these dynamical systems.
- The research proves that almost every number has multiple double-base expansions while unique expansions form a set of Lebesgue measure zero.
Invariant Measures and Dynamical Properties in Double Base Expansions
Overview
This paper addresses the ergodic theoretic and measure-theoretic properties of expansions of real numbers in double-base systems, or (q0,q1)-expansions, where q0,q1>1 and q0+q1≥q0q1. The focus is on the dynamical systems generated by the greedy and lazy algorithms—canonical representatives of maximal and minimal lexicographical expansions—in this context. The main results demonstrate the existence, explicit structure, and statistical properties of unique absolutely continuous invariant probability measures (ACIPMs) for the greedy and lazy maps, their equivalence to Lebesgue measure on specified intervals, and the exactness of the corresponding dynamical systems. Theoretical implications for the uniqueness of expansions are also derived, connecting to the classical theory of q-expansions in single and multiple bases.
Double-Base Expansions and Dynamical Systems
Let Q=(q0,q1), and for x∈IQ=[0,1/(q1−1)], a Q-expansion is a sequence (ci)∈{0,1}N such that
x=i=1∑∞qc1⋯qcici.
Assuming q0+q1≥q0q1, every q0,q1>10 has at least one such expansion, and the set q0,q1>11 is invariant under the relevant transformations. The greedy map q0,q1>12 and lazy map q0,q1>13 are defined as piecewise-affine transformations derived from the digit selection mechanisms of the corresponding expansion algorithms.
Absolutely Continuous Invariant Measures
A central result establishes that both the greedy and lazy maps admit unique ACIPMs, denoted q0,q1>14 and q0,q1>15, which are mutually singular in general and are each equivalent to Lebesgue measure on their supports: q0,q1>16 for the greedy map (q0,q1>17), and q0,q1>18 for the lazy map (q0,q1>19). This is rigorously obtained by conjugating the interval q0+q1≥q0q10 to q0+q1≥q0q11, employing the Lasota-Yorke framework for piecewise expanding maps [18,19], and invoking Perron-Frobenius theory to yield explicit jump function formulas for invariant densities. The densities are shown to be bounded, with q0+q1≥q0q12 non-increasing and q0+q1≥q0q13 non-decreasing.
Exactness and Strong Stochasticity
Both the greedy and lazy dynamical systems are shown to be exact, and thus also mixing and ergodic in the measure-theoretic sense. Exactness implies rapid loss of memory of initial conditions and maximal stochasticity with respect to the invariant measure, solidifying the connection between digit expansion in double base systems and strong random-like dynamical behavior.
Properties of Unique Expansions
An important theoretical implication is the measure-theoretic size of the set of numbers with unique q0+q1≥q0q14-expansions. Under the strict inequality q0+q1≥q0q15, this univoque set q0+q1≥q0q16 has Lebesgue measure zero, paralleling the single-base case. Moreover, almost every q0+q1≥q0q17 admits a continuum of distinct q0+q1≥q0q18-expansions, a fact deduced via ergodic and Birkhoff-type limit theorems for the q0+q1≥q0q19-expansion process.
Explicit Invariant Density Representation
Proportional explicit representations for the invariant density are given in terms of jump functions involving the greedy and lazy expansions of critical points (the endpoints of q0). The density for the greedy algorithm, q1, is expressed as a sum over partitions determined by the expansion structure, and similarly for q2. In certain degenerate cases (e.g., q3), these densities are constant and coincide with the normalized Lebesgue density.
Theoretical and Practical Implications
From a theoretical perspective, these results provide a comprehensive framework for understanding measure-theoretic and dynamical properties in multi-base and non-uniform numeration systems, extending classical results by Rényi, Parry, and Lasota-Yorke to double-base contexts. The exactness and ACIPM structure are essential for statistical properties of digit sequences, dimension theory, and further arithmetic applications, such as normality and entropy calculations.
Practically, these findings inform algorithmic construction of random and pseudo-random number generators, numerical systems with redundancy, and coding schemes, especially where non-integer or multiple bases arise (such as in fractal geometry or non-uniform quantization).
Conclusion
This paper rigorously characterizes the invariant measure structure and strong stochastic properties of digit expansion algorithms in double base numeration systems. It establishes the existence and explicit structure of unique ACIPMs for both greedy and lazy maps, the exactness of the induced dynamical systems, and proves that unique expansions are a null phenomenon in this context. These results solidify the connection between symbolic dynamics, ergodic theory, and number theory for double-base representations, suggesting further exploration into higher-dimensional or more general non-uniform numeration schemes and their associated dynamical systems.
References:
- Lasota, A., Yorke, J. A., 1982. "Exact dynamical systems and the Frobenius-Perron operator." Trans. Amer. Math. Soc. 273, no. 1: 375-384.
- Parry, W., 1960. "On the β-expansions of real numbers." Acta Math. Acad. Sci. Hungar. 11: 401-416.
- Rényi, A., 1957. "Representations for real numbers and their ergodic properties." Acta Math. Acad. Sci. Hungar. 8: 477-493.
See "Invariant measure for double base expansions" (2605.08641).