Drag law. The default is the Schiller-Naumann correction to Stokes drag, valid up to Re about 800: f(Re) = 1 + 0.15·Re0.687, Re = 2·r·ρw·|u|/μ with |u| the grain's speed relative to the water, so the relaxation rate becomes k·f(Re). Pure Stokes (slider to 0) is exact only for Re < 1, but a 0.5 mm gold grain sinks at Re of order 100, where plain Stokes over-predicts the sink speed by a factor of several; the readout table shows Re for the median grain of each material. Each time step uses the exact exponential relaxation toward the local terminal velocity, so it is stable for any k·dt.
Every grain here obeys Stokes' law: the viscous drag on a small sphere settling in water gives a settling (relaxation) rate k = 4.5·μ/(ρ·r²) and an equilibrium sink velocity v = (2/9)·(ρ−ρw)·g·r²/μ. Gold's specific gravity (19.3) against water is nearly 8× that of quartz sand (2.65), so a gold grain accelerates toward the bed far faster than a sand grain of the same size — the entire principle behind panning and sluicing.
| Material | Specific gravity |
| 🟡 Gold | 19.3 |
| ⬛ Black sand (magnetite) | 5.2 |
| ⚪ Quartz sand | 2.65 |
| 🟤 Clay / silt | 1.8 |
Downstream of each riffle bar the flow recirculates, dropping the near-bed water speed — a real low-velocity eddy pocket. A grain only gets trapped there if it has settled to the bed and is moving slower than the pocket can re-entrain it, so taller riffles (bigger pockets) trap more but also silt up with worthless sand; too much water flow sweeps everything through; too little separates nothing. Agitation adds turbulent velocity fluctuations on top of the mean flow — real panning shake — which lofts light, fine grains back into suspension while dense gold stays put (differential mobility, the same physics as kinetic sieving). Motion here runs at roughly 1/8 real speed so the settling — which is genuinely faster than a video frame for a 300 µm gold grain — stays visible.
- Bed contact — near-bed speed follows a boundary-layer profile u(y) = U(1 - e^(-y/1.5 cm)) sampled at a 0.3 cm roughness height; a grain resting on the bed obeys Coulomb friction (mu = 0.08 on its submerged weight), so dense gold stays put where light sand slides on. A grain counts as recovered only if it sits in a riffle pocket nearly at rest (below 1 cm/s) for 0.35 s.
- Mass bookkeeping — fed = kept + lost in tailings + still in the water (+ grains dropped by the 260-grain display cap, which are removed from "fed"); the check below verifies it to rounding error.
- Simplifications — 2D side view of a 3D flow, uniform mean flow with a single-scale random turbulence, spheres (real gold is flaky and settles slower), no grain-grain collisions or hindered settling, riffle eddy modelled as a rectangular pocket with a reduced mean velocity, capture rule is empirical. It teaches the ranking and trends, not absolute recoveries of a real sluice.
- Recovery rate — mass of gold trapped behind riffles ÷ mass of gold fed in.
- Concentrate grade — gold's mass share of everything trapped; a well-tuned sluice raises this far above the feed's own gold fraction (2% by count here, much less by mass since gold grains are the smallest).