Modelling aerosol-based exposure to SARS-CoV-2 by an agent based Monte Carlo method: Risk estimates in a shop and bar.

Present day risk assessment on the spreading of airborne viruses is often based on the classical Wells-Riley model assuming immediate mixing of the aerosol into the studied environment. Here, we improve on this approach and the underlying assumptions by modeling the space-time dependency of the aerosol concentration via a transport equation with a dynamic source term introduced by the infected individual(s). In the present agent-based methodology, we study the viral aerosol inhalation exposure risk in two scenarios including a low/high risk scenario of a "supermarket"/"bar". The model takes into account typical behavioral patterns for determining the rules of motion for the agents. We solve a diffusion model for aerosol concentration in the prescribed environments in order to account for local exposure to aerosol inhalation. We assess the infection risk using the Wells-Riley model formula using a space-time dependent aerosol concentration. The results are compared against the classical Wells-Riley model. The results indicate features that explain individual cases of high risk with repeated sampling of a heterogeneous environment occupied by non-equilibrium concentration clouds. An example is the relative frequency of cases that might be called superspreading events depending on the model parameters. A simple interpretation is that averages of infection risk are often misleading. They also point out and explain the qualitative and quantitative difference between the two cases-shopping is typically safer for a single individual person.

Modelling aerosol transport and virus exposure with numerical simulations in relation to SARS-CoV-2 transmission by inhalation indoors.

We provide research findings on the physics of aerosol and droplet dispersion relevant to the hypothesized aerosol transmission of SARS-CoV-2 during the current pandemic. We utilize physics-based modeling at different levels of complexity, along with previous literature on coronaviruses, to investigate the possibility of airborne transmission. The previous literature, our 0D-3D simulations by various physics-based models, and theoretical calculations, indicate that the typical size range of speech and cough originated droplets ( d ⩽ 20 µ m ) allows lingering in the air for O ( 1 h ) so that they could be inhaled. Consistent with the previous literature, numerical evidence on the rapid drying process of even large droplets, up to sizes O ( 100 µ m ) , into droplet nuclei/aerosols is provided. Based on the literature and the public media sources, we provide evidence that the individuals, who have been tested positive on COVID-19, could have been exposed to aerosols/droplet nuclei by inhaling them in significant numbers e.g. O ( 100 ) . By 3D scale-resolving computational fluid dynamics (CFD) simulations, we give various examples on the transport and dilution of aerosols ( d ⩽ 20 µ m ) over distances O ( 10 m ) in generic environments. We study susceptible and infected individuals in generic public places by Monte-Carlo modelling. The developed model takes into account the locally varying aerosol concentration levels which the susceptible accumulate via inhalation. The introduced concept, 'exposure time' to virus containing aerosols is proposed to complement the traditional 'safety distance' thinking. We show that the exposure time to inhale O ( 100 ) aerosols could range from O ( 1 s ) to O ( 1 min ) or even to O ( 1 h ) depending on the situation. The Monte-Carlo simulations, along with the theory, provide clear quantitative insight to the exposure time in different public indoor environments.

Machine learning plastic deformation of crystals.

Plastic deformation of micron-scale crystalline solids exhibits stress-strain curves with significant sample-to-sample variations. It is a pertinent question if this variability is purely random or to some extent predictable. Here we show, by employing machine learning techniques such as regression neural networks and support vector machines that deformation predictability evolves with strain and crystal size. Using data from discrete dislocations dynamics simulations, the machine learning models are trained to infer the mapping from features of the pre-existing dislocation configuration to the stress-strain curves. The predictability vs strain relation is non-monotonic and exhibits a system size effect: larger systems are more predictable. Stochastic deformation avalanches give rise to fundamental limits of deformation predictability for intermediate strains. However, the large-strain deformation dynamics of the samples can be predicted surprisingly well.