Model. The room (V = 168 m³) is treated as one well-mixed box. Infectious "quanta" (Wells-Riley: one quantum is the dose that infects 63 % of susceptibles) are emitted at S = I q (1 − ηe) per hour and removed at a total first-order rate λ = ACH + CADR/V + ksettle + kinact. Then dC/dt = S/V − λC, so C(t) = S/(Vλ) (1 − e−λt).
Risk. A susceptible breathing p (1 − ηi) m³/h inhales dose D = p(1−ηi)∫C dt and is infected with probability P = 1 − e−D. With only ventilation and a long stay this reduces to the Rudnick-Milton form D = I q p t / Q. Expected new cases = (N − I) P.
CO₂. dc/dt = N g/V − ACH (c − cout); steady state cout + 1000 N g / Q ppm. CO₂ is a ventilation proxy only: HEPA filters remove aerosol but not CO₂.
The cloud. Each orange dot is created at the infectious student's mouth at a Poisson rate and removed with hazard λ, so the count obeys the same ODE as the box model (chart). Its drift is an illustrative convection cell (supply at the left ceiling, exhaust at the right) plus a random walk, not CFD; real rooms have local plumes, so the box model is a room average, not a seat-by-seat prediction. Tap a student to make them the infectious one.
Illustrative parameters (not measurements). Quanta rates 2 / 10 / 40 per hour (breathing / talking / loud), breathing 0.4 / 0.5 / 0.7 m³/h, mask efficiencies, HEPA 250 m³/h per unit, ksettle = 0.24 /h are round figures in the range of published studies; kinact = 0.63 /h matches the ~1.1 h aerosol half-life reported for SARS-CoV-2 (van Doremalen et al., 2020). Real values vary by an order of magnitude. This is a physics illustration, not medical or health advice.
Honest simplifications. Perfect mixing, constant emission, no re-suspension, identical susceptibles, no immunity, no door traffic, one efficiency per mask type; the teacher is not counted. Figures are drawn larger than a to-scale 8 m room would show.
Independent checks (run on load, same core as the 3D page):