The Weierstrass function is a famous mathematical object constructed by adding together infinitely many oscillating waves, where each successive wave vibrates much faster than the previous one (this "lacunary" property means the frequencies grow rapidly). The resulting function is continuous everywhere but jagged at every scale, making it a classic example of a fractal-like curve. This paper studies how the total amount of up-and-down variation in such functions can be measured and characterized, using a generalization where the basic wave shape is not necessarily a cosine or sine, but any sufficiently smooth periodic function whose Fourier representation satisfies a specific gap condition on its frequency content.
The central quantity of interest is called the variation function, which tracks how much a function fluctuates when you zoom in at finer and finer scales using a systematic grid. The authors show that, under their generalized construction, this variation function behaves in a predictable and well-characterized way. The key mathematical tool they use is a martingale central limit theorem, which is a result from probability theory that describes how certain cumulative sums of dependent random-like quantities converge to a normal bell-curve distribution. This probabilistic framing, even though the functions themselves are deterministic, turns out to be a powerful way to understand the aggregate oscillation behavior.
The results extend and generalize earlier work by Han, Schied, and Zhang from 2021, broadening the class of functions for which such variation results are known. A notable feature is the flexibility of the framework: the periodic building-block function can itself be something exotic, like the Takagi function (another well-known fractal construction) or even another lacunary series. This opens the door to studying a wide zoo of irregular functions within a unified analytical framework, with potential relevance to fields like mathematical finance, where such rough and irregular functions appear naturally in modeling price paths.