More than a year ago, research from the Finnish supercomputing center CSC went viral – so to speak – when their simulations showed how viral particles from a cough could spread through and linger across the aisles of a grocery store. New research led by Michel Boufadel and Fangda Cui from the New Jersey Institute of Technology, however, has used the Comet system at the San Diego Supercomputer Center (SDSC) to demonstrate that grocery stores, if not COVID-safe, might at least be less of a viral warzone than previously feared.
The researchers used Comet – which delivers 2.76 peak petaflops via 1,944 Intel Haswell CPU nodes 72 Nvidia GPU nodes – to simulate viral particles expelled among grocery store isles at a series of time intervals: ten seconds after release, 100 seconds, 200 seconds and five minutes. Furthermore, the researchers stratified the study based on the stickiness of those particles.
“We used allocations from the National Science Foundation’s Extreme Science and Engineering Discovery Environment [XSEDE] for these Comet simulations that helped us investigate the transport of virus-laden particles in an archetypical supermarket 40 meters long by 30 meters wide by 4.5 meters (ceiling height),” elaborated Boufadel, who is a professor of civil and environmental engineering at NJIT, in an interview with SDSC’s Kimberly Bruch and Cynthia Dillon. “We considered three efficiency situations of attachment on surfaces: zero percent, 25 percent and 100 percent attachment. For example, the 25 percent attachment efficiency means that out of 100 particles that touch the surface, only 25 percent attached to a surface – thus zero percent means no attachment.”
The results were startling: despite the vastly disparate stickiness of the particles, the 25 percent and 100 percent levels had the same “efficiency” of attachment as they collided with the shelves, floors and ceilings of the simulated spaces. Overall, those sticky impacts with surfaces reduced the concentration of airborne particles by as much as half, leading Boufadel to cite display shelves in the aisles as a helpful measure for preventing the spread of COVID in grocery stores (as long as visitors are still using hand sanitizer).
Now, the team is moving ahead to simulating even more minute interactions, such as those at the micron level, using additional XSEDE computing resources. “The goal was a quantification, and the superior ability of XSEDE allowed us to have a high confidence in the results,” Boufadel said. “XSEDE has opened a new universe for us.”
To learn more about this research, read the coverage from SDSC’s Kimberly Bruch and Cynthia Dillon.