A characterization of Benford's law in discrete-time linear systems (Q300362)

The study of digits generated by dynamical processes is a classical subject that continues to attract interest from many disciplines, including ergodic and number theory, analysis and statistics. A recurring theme across the disciplines is the surprising ubiquity of a logarithmic distribution of digits often referred to as \textit{Benford's Law} (BL). The most well-known special case of BL is the so-called (decimal) first-digit law which asserts that \[ \mathbb{P}\left(\text{leading digit}_{10}=d_{1}\right)= \log_{10}\left(1+\frac{1}{d_{1}}\right) \qquad \left(d_{1}=1, \ldots, 9\right), \] where \(\text{leading digit}_{10}\) refers to the leading (or first significant) decimal digit, and \(\log_{10}\) is the base-\(10\) logarithm. The main purpose of this paper is to investigate the following problem: Given \(d\in \mathbb{N}\) and real numbers \(a_{1}, a_{2}, \ldots, a_{d}\) with \(a_{d}\neq 0\), consider the following autonomous, \(d^{\text{th}}\) order linear difference equation \[ x_{n}=a_{1}x_{n-1}+a_{2}x_{n-2}+\cdots+a_{d}x_{n-d} \qquad \left(n\in \mathbb{N}, n\geq d+1\right). \tag{1} \] Under which conditions on \(a_{1}, \ldots, a_{d}\) and presumably also on the initial values \(x_{1}, \ldots,\allowbreak x_{d}\) does the solution \((x_{n})_{n\in \mathbb{N}}\) satisfy \[ \lim_{N\to \infty}\dfrac{\#\left(n\leq N\, | \, \text{leading digit}_{10}(x_{n})=d_{1}\right)}{N}=\log_{10}\left(1+\frac{1}{d_{1}}\right) \qquad \left(n\in \mathbb{N}, n\geq d+1\right). \tag{2} \] The result below (which is Theorem 3.16 in the paper) provides a necessary and sufficient condition for every nontrivial solution of (1) to satisfy (2). Theorem. Let \(a_{1}, \ldots, a_{d}\) be real numbers with \(a_{d}\neq 0\) and \(b\in \mathbb{N}\setminus \left\{ 1 \right\}\). Then the following assertions are equivalent. {\parindent=0.7cm\begin{itemize}\item[--] Every solution \((x_{n})_{n\in \mathbb{N}}\) of (1) is \(n\)-Benford unless \(x_{i}=0\) for all \(i=1, \ldots, d\). \item[--] With the polynomial \[ p(z)=z^{d}-a_{1}z^{d-1}-\cdots-a_{d-1}z-a_{d} \] the set \[ \left\{z\in \mathbb{C}\, | \, p(z)=0\right\} \] is \(n\)-nonresonant. \end{itemize}}