Average orders and the Erdős–Kac viewpoint

Many arithmetic functions are “wild” pointwise but have predictable average behavior. This page gives a short refresher on average order, normal order, and the Erdős–Kac phenomenon. See [Tenenbaum, 2015] and [Erdős and Kac, 1940].

Average order vs. normal order

  • Average order: a function \(A(x)\) such that

    \[ \sum_{n\le x} f(n) \approx \sum_{n\le x} A(n) \]
    or \(f\) has mean value described by \(A\).

  • Normal order: a function \(N(n)\) such that “most” integers satisfy

    \[ f(n) \approx N(n). \]

A classic example: for most \(n\), \(\omega(n)\) is close to \(\log\log n\).

Erdős–Kac theorem (informal statement)

Let \(\omega(n)\) be the number of distinct prime factors. Erdős and Kac proved that the normalized variable

\[ \frac{\omega(n)-\log\log n}{\sqrt{\log\log n}} \]

has an approximately standard normal distribution over \(n\le x\) as \(x\to\infty\). [Erdős and Kac, 1940]

Experiment ideas

  • for a chosen \(N\), compute \(\omega(n)\) for \(n\le N\) and histogram the normalized values

  • compare the empirical mean/variance with \(\log\log N\)

  • explore how the fit improves as \(N\) grows

Experiments in this repository

  • E094 — Erdős–Kac side-by-side for ω and Ω (normal order / scaling).