The top chart integrates the mean-field kinetics of the same reaction ladder used industrially:
TG --k--> DG --k--> MG --k--> GL (+FAME at every step)
Instead of stepping 150 individual particles with a random dice-roll each frame (the discrete Monte-Carlo approach), this engine integrates the deterministic ODE system directly with adaptive sub-stepping:
d[TG]/dt = -k[TG]
d[DG]/dt = k[TG] - k[DG]
d[MG]/dt = k[DG] - k[MG]
d[GL]/dt = k[MG]
with the same Arrhenius rate law k = A·exp(−Ea/RT)·(catalyst/0.75) and the same methanol-ratio equilibrium cap that halts progress once the aggregate FAME yield reaches its ceiling. Dots in the reacting band are recolored in rank order as the fractions evolve, so the population you see always matches the integrated concentrations exactly.
The bottom column is a genuinely separate physics layer with no equivalent in the 3D version, which only teleports particles between fixed height bands. Here, every FAME and glycerol droplet spawned by the reaction gets a real terminal velocity from Stokes' law:
v = (2/9)·(ρ_particle − ρ_medium)·g·r² / μ(T)
Glycerol (≈1260 kg/m³) is far denser than the mixture (≈900 kg/m³) and sinks quickly; biodiesel (≈880 kg/m³) is only slightly lighter and rises slowly — the panel's live ratio readout shows exactly how much faster glycerol settles. Viscosity μ falls with temperature (Andrade-type law), so raising the temperature slider speeds up separation as well as the reaction. The turbulence slider adds random drift that fights the settling drift, modeling incomplete mixing shutdown.
Drag across the top chart to drop a time cursor and read exact fractions at any past moment.