Eclogite Garnet Separation: Heavy-Liquid Centrifuge, Stokes Settling and the Density Cut

Spin crushed eclogite in a heavy liquid: garnet sinks, lighter omphacite floats. Tune the liquid density, speed and grain size and watch the cut sharpen or blur.

Switch to 2D

Presets

Playback

Parameters

Camera

Layers

Quality

Model check

The science

What is modelled

Eclogite is a dense high-pressure rock made mainly of red garnet and green omphacite. Geologists separate the minerals for dating and chemistry by suspending crushed grains in a heavy liquid and spinning it in a centrifuge: grains denser than the liquid are flung to the bottom of the tube, lighter ones stay at the surface. This page models a four-tube swing-bucket centrifuge. Each grain has its own density and size, and moves under centrifugal acceleration against Stokes drag.

Equations

  • Centrifugal acceleration at radius r: a = omega^2 r with omega = rpm x pi/30; the relative force is omega^2 r / g (in g).
  • Stokes settling, valid for small Reynolds numbers: v = d^2 (rho_p - rho_f) a / (18 mu). With a = omega^2 r this gives dr/dt = k r, k = d^2 (rho_p - rho_f) omega^2 / (18 mu) and the exact solution r(t) = r0 e^(k t).
  • Crowding: hindered settling (1 - phi)^4.65 (Richardson and Zaki, low Reynolds number).
  • Time to reach a wall: t = ln(R / r0) / |k|; a grain sinks if rho_p > rho_f, otherwise it floats to the meniscus.
  • Density cut: each mineral has a spread of densities (normal with mean and standard deviation, see list); the fraction that sinks is 1 - Phi((rho_f - mean)/sd). Garnet 3.90 +/- 0.18, omphacite 3.32 +/- 0.06, rutile 4.25 +/- 0.03, quartz 2.65 +/- 0.01 g/cm3 are typical literature values (garnet varies with its almandine, pyrope and grossular content).

Checks run in the page

  • A fourth-order Runge-Kutta integration of dr/dt = k r matches the closed form r0 e^(kt) to 1e-6.
  • The Stokes speed of a 100 um quartz grain in water at 1 g matches the hand calculation (about 9 mm/s).
  • The density-cut formula agrees with counting 20 000 sampled garnet grains.
  • After a long run each grain sits at the wall predicted by the sign of its density difference (0 mismatches).

Honest simplifications

  • The viscosity law mu = 1 mPa s x e^(1.6 (rho_f - 1)) is an illustrative fit for tungstate-type liquids, not a data sheet; real viscosity also depends on temperature and the liquid used. Several real heavy liquids are toxic, so this page is about the physics, not a laboratory recipe.
  • Grains are treated as spheres with one density each (no inclusions, no shape factor); the real separation is also limited by mineral intergrowths.
  • The rotor accelerates exponentially (time constant 2.5 s), the tube is a straight radial channel and Coriolis and wall effects are ignored. The largest Reynolds number is reported; above 1 Stokes drag is too optimistic.
  • Grains are drawn about 20 times larger than true so they can be seen; the rotor turns slower on screen than the real one.

What to try

  • Sweep the liquid density in graph 2: garnet recovery rises while sink purity falls as omphacite also starts to sink near 3.3 to 3.5.
  • Halve the grain size: the time to a wall grows with 1/d^2 (graph 3).
  • Try 300 rpm: the acceleration is about 10 g, so the run is slow.
Loading model...
Loading Blender model...
Radial position of one grain of each mineral against time
Density cut: garnet recovery and purity against liquid density
Time to reach a wall against grain diameter
Made with 🟠 Blender · Three.js