Compression and information entropy of binary strings from the collision history of three hard balls

TL;DR

This paper explores how to measure the entropy of a chaotic system of three hard spheres on a ring by transforming collision histories into binary strings and applying various randomness tests and compression algorithms.

Contribution

It introduces a novel method that does not rely on the ergodic hypothesis to analyze the entropy and randomness of collision-generated binary strings from a chaotic system.

Findings

Shannon entropy cannot distinguish random from deterministic strings.

Diehard tests sometimes misidentify collision strings as random.

Compression algorithms effectively detect non-randomness and low information content.

Abstract

We investigate how to measure and define the entropy of a simple chaotic system, three hard spheres on a ring. A novel approach is presented, which does not assume the ergodic hypothesis. It consists of transforming the particles collision history into a sequence of binary digits. We then investigate three approaches which should demonstrate the non-randomness of these collision-generated strings compared with random number generator created strings: Shannon entropy, diehard randomness tests and compression percentage. We show that the Shannon information entropy is unable to distinguish random from deterministic strings. The Diehard test performs better, but for certain mass-ratios the collision-generated strings are misidentified as random with high confidence. The zlib and bz2 compression algorithms are efficient at detecting non-randomness and low information content, with compression efficiencies that tend to 100% in the limit of infinite strings. Thus compression algorithm entropy is non-extensive for this chaotic system, in marked contrast to the extensive entropy determined from phase-space integrals by assuming ergodicity.