Respiratory droplets have been at the centre of public attention since the COVID-19 pandemic, yet we still know surprisingly little about the mechanisms that shape their size distributions. Most studies stop at finding a “good fit” with a lognormal or Weibull curve. My recent article takes a different route: it asks what we learn if we treat human aerosols not just as droplets in air, but as carriers of information about the underlying fluid-dynamic processes that created them.
At the core of the work is an infodynamic viewpoint: every aerosol size distribution has an “informational richness” that can be quantified. Using the notion of differential informature, the study measures how much information is needed to describe a given distribution—its infodiversity. Different mechanisms of droplet formation (film bursting in the deep lung, ligament breakup in the airway, turbulent stripping in the upper airways) leave different informational fingerprints.
To expose these fingerprints, the paper revisits classic and recent datasets of human aerosols from breathing, speaking, coughing and even sneezing, measured with a variety of instruments (SMPS, APS, IMI, OPS). It first puts all data on the same footing, as proper probability density distributions, and reconstructs them with a refined, Fibonacci-based binning that preserves detail without inflating information artificially. Then it compares the empirical distributions with mechanistic candidates: Lognormal, Gamma and Weibull PDFs and their mixtures, evaluated not only by geometric goodness-of-fit, but by a new composite normalized infodynamic gap (CNIG) that demands both small information loss and physical plausibility.
The overall picture is simple: human respiratory aerosols are almost never created by a single process. In the 22 situations analyzed, the data are best explained by a combination of three basic patterns of droplet formation. Roughly two thirds of the contribution comes from one pattern, linked to the bursting of thin liquid films in the airways. About one fifth comes from a second pattern, associated with step-by-step break-up of droplets in chaotic airflow. The remaining share is tied to the fragmentation of liquid “threads” that form and snap as we breathe, speak or cough. The variety of droplet sizes – what we measure as the “informational richness” of the aerosol – spans a wide range and depends not only on what the person is doing, but also on how we measure it. Some instruments miss the smallest droplets and, in doing so, hide part of the story that the aerosol is trying to tell.
Beyond these specific numbers, the article’s main contribution is conceptual. It offers a model–physics convergence framework where a PDF is acceptable not just because it fits the data, but because its information content matches what the underlying physics can plausibly generate. New metrics such as the Polarity Index summarize how much a given aerosol is governed by film bursting, ligament fragmentation or turbulent extraction. This, in turn, helps identify which datasets are most “informationally rich” and therefore most suitable as inputs for simulations of infection risk, ventilation design or mask performance.
Perhaps the most intriguing aspect is the broader message: respiratory aerosols become a window into the fluid dynamics of life. By reading their informational structure, we can infer how living systems fragment, transport and reorganize liquid at multiple scales. The work suggests future pathways—incorporating size–velocity correlations, developing real-time infodynamic diagnostics, and extending the approach to other biological aerosols. It invites us to see each exhaled cloud not only as a health hazard, but as a complex, information-bearing signature of the organism that produced it.
See paper at https://doi.org/10.1063/5.0293235