[Paper Review] The Limited Scaling Range of Empirical Fractals

David Avnir, Ofer Biham+2 more|arXiv (Cornell University)|Jan 7, 1998

Theoretical and Computational Physics1 references248 citations

TL;DR

This paper challenges the widespread use of 'fractal' to describe empirical data, arguing that most reported fractal scaling in physical systems spans only 0.5–2.0 decades—far short of the many orders of magnitude required for true fractality. The authors advocate for retaining power-law analysis under Eq. (1) for its practical utility, even if the term 'fractal' is misleading in the strict mathematical sense.

ABSTRACT

The notion of the abundance of fractals is critically re-examined in light of surprising data regarding the scaling range in empirical reports on fractality.

Motivation & Objective

• To challenge the common practice of labeling empirical systems as 'fractal' based on limited scaling ranges.
• To investigate whether power-law scaling over short ranges (less than one order of magnitude) can legitimately be interpreted as fractal.
• To assess the scientific utility of fractal terminology in experimental physics despite its conceptual inaccuracy.
• To examine whether the label 'fractal' is still justifiable given that most experimental data spans only 0.5–2.0 decades of scaling.
• To argue that the real value lies in multiple-resolution analysis, not in the fractal label per se.

Proposed method

• Analyzing 96 experimental papers from Physical Review journals (1990–1996) that reported fractal analysis.
• Extracting and quantifying the number of decades of power-law scaling used to claim fractality.
• Using log-log plots of property P versus resolution r to fit Eq. (1): P = k·r^{f(D)}
• Constructing a histogram of scaling ranges to assess the distribution of reported fractal scaling extents.
• Comparing empirical scaling ranges to the theoretical requirement of infinite scaling for true fractals.
• Evaluating the visual and conceptual appeal of self-similarity in limited-range power laws.

Experimental results

Research questions

• RQ1 Is the term 'fractal' scientifically justified when applied to empirical systems with scaling ranges of less than one order of magnitude?
• RQ2 To what extent do experimental results in physical systems truly exhibit the infinite scaling required for mathematical fractals?
• RQ3 What is the practical value of power-law analysis in experimental systems, independent of the fractal label?
• RQ4 Why has the term 'fractal' become entrenched in scientific literature despite its inaccuracy in most empirical cases?
• RQ5 Can the benefits of multiple-resolution analysis be preserved without relying on the concept of fractality?

Key findings

• The majority of reported experimental fractality is based on scaling ranges of only 0.5 to 2.0 decades, with a peak at 1.3 decades.
• True mathematical fractals require infinitely many orders of magnitude of scaling, which is not observed in empirical data.
• Even two iterations of a Koch curve (one order of magnitude) do not constitute a fractal object in the strict sense.
• The limited scaling range is primarily due to physical cutoffs: lower cutoffs from basic structural units and upper cutoffs from system size or physical constraints.
• Despite the lack of rigorous fractality, power-law fitting (Eq. 1) remains useful for condensing complex geometry and enabling structure-property correlations.
• The continued use of the term 'fractal' may be justified not by mathematical accuracy but by its symbolic value and entrenched usage in scientific practice.