Statistical measure of dispersity in complex weighted networks of low temperature plasma chemistry

Abstract A chemical reaction system is frequently complex when its molecular species are excited by high-energy external sources. Here we examine molecular (SiH 4 ) plasma chemistry in which high-energy electrons dissociate, decompose and polymerize many molecules in parallel. Its network statistics have been reported to exhibit complexity with scale-freeness, but its analytical model was too simple on the assumption of equally-weighted edges in the reaction network. In this study, using weighted networks, we quantitatively investigate a complex chemical reaction system by Shannon entropy and conditional entropy, which can be derived by probability distributions of constituent species and reaction rates. Quantities of Shannon entropy increase as the reactions proceed, implying increased broadening spectra of species, while those of conditional entropy are not always the case. With additional hypothetical attempts of arbitrary removal of fractional species, these results revealed in this study suggest that complexity properties that include species dispersity and system robustness can be simply quantified by entropy analysis, and a machine-learning model applicable to network topology coincides with the numerical results of this complex chemical system.