OPOpenPithMake curiosity visibleChemistry lab / Crystal casting workshop

About this lab

One real casting session, turned into four experiments you can test.

These four experiments come from one real casting session: two cans of slow-cure resin marked A3:B1, a kitchen scale, a rotary vane vacuum pump, a few white silicone cluster moulds, and the results: milky clusters and cylinders with a specimen inside. This resin has no data sheet. So the four experiments use only principles that do not depend on product values: rounding error, Boyle’s law, surface scattering and Archimedes’ principle.

The principles of curing, heat and bubbles are in the crystal cluster model lab. This page covers four problems you meet at the bench. Values marked illustrative show trends only and are not measurements.

Four findings

Four findings

  • Weighing error: the ratio is 3 : 1 by mass, so a 40 g batch needs 30 g of A and 10 g of B. With a 1 g division, the worst-case ratio error is 11.9 %. For a 20 g batch it is 24.1 %, and for 100 g it falls to 4.7 %. With a 0.1 g division, a 20 g batch gives only 2.3 %. For a 5 % tolerance the batch must be at least 94 divisions. For 2 % it must be 234 divisions.
  • Vacuum degassing: at 10 kPa absolute, a bubble grows to 10.1 times its volume and 2.16 times its diameter, and rises 4.7 times faster. At 3 kPa the factors are 33.8, 3.23 and 10.4. With 4 % gas in the resin, half a cup rises to 68 % of the cup height at 10 kPa, and to 116 % at 3 kPa, which is over the rim. At 3 kPa the fill must stay under 43 % of the cup height. A quarter cup of thin resin held for 5 minutes keeps 81 bubbles with no vacuum, 34 at 10 kPa and 12 at 3 kPa. With resin ten times thicker, 79 are still left at 3 kPa.
  • Sanding and polishing: the demoulded face has a roughness of 250 nm and is only 0.5 % clear. When wet it is 41.0 % clear, so the haze is on the surface. After P400 the roughness is 583 nm. Step by step to P2500 it is 146 nm, and after compound it is 17 nm: 88.5 % clear, against a limit of 90.5 %. Skipping from 400 straight to 2500 leaves 411 nm, still matte. With micro-bubbles inside, the wet piece is only 14.4 % clear, and the finest finish gives only 31.0 %.
  • Embedding a specimen: the cylinder is 5 cm deep inside and the specimen is 2 cm tall. Quartz (2.65 g/cm³) and sandstone (2.2 g/cm³) sink, with the centre 1.5 cm too low. Pumice (0.64 g/cm³) is only 59 % submerged in EpoxAcast 692. A single 5 cm pour of EpoxAcast 690 peaks near 166 °C in the model. For 692 it is about 40 °C, above its heat deflection temperature of 38 °C. With layers, 690 needs 6 pours and 692 needs 2, and both take 144 hours. Pumice needs one more pour to fix it in place. Sandstone in a layered pour still releases about 225 bubbles, and none after sealing. Temperatures and bubble counts are illustrative.

Model and sources

Model and sources

  • Weighing: weigh A first, then add B in the same cup. Each reading is rounded by up to half a division, so B can be wrong by one full division d. The worst-case ratio factor is 3(B + d) / (A − d/2). This gives the smallest batch M = d(14 + 2τ) / (3τ), where τ is the tolerance. The 240 repeat weighings use a seeded random number, with the true mass spread evenly inside the rounding interval. The 2 % and 5 % tolerances are illustrative.
  • Vacuum: Boyle’s law p·V = constant, with standard atmospheric pressure 101.325 kPa. Rising speed follows Stokes’ law and goes with the square of the diameter. Foam height = fill level × (1 − φ + φ·p₀/p). These values are illustrative: gas content φ of 4 %, viscosity of 0.3 and 3 Pa·s, density of 1.1 g/cm³, a 5-minute hold, and 160 bubbles with a median diameter of 0.12 mm. The model leaves out dissolved gas that comes out of solution and resin that evaporates at low pressure. Bubbles are drawn larger than life.
  • Sanding: grain sizes are the median values of the FEPA P grades. P400 is 35.0 µm, P800 is 21.8 µm, P1200 is 15.3 µm and P2500 is 8.4 µm. The clear transmission of a rough interface is exp(−(2π·σ·Δn/λ)²) with λ = 550 nm. The Fresnel transmission of each interface is 1 − ((n₁ − n₂)/(n₁ + n₂))². Light crosses two faces. The resin index is that of 690, 1.565, and water is 1.333. These values are illustrative: σ = grain size ÷ 60, 250 nm for the demoulded face, 1 µm for the compound, the rule for scratches left after skipping, and 35 % transmission for micro-bubbles inside.
  • Embedding: Archimedes’ principle, submerged fraction = specimen density ÷ resin density. The peak temperature of each pour comes from the heat balance model of the “Depth and heat” lesson in the crystal cluster model lab. Pore gas expands at constant pressure, and the share pushed out is 1 − T_room/T_peak in kelvin. Specimen density and porosity are typical values. These values are illustrative: resin wetting pushes out one tenth of the pore gas, bubbles are 1 mm across, and the sealer closes every pore. The cylinder mould is drawn at twice its size.

These four experiments do not replace three things you need before real work: the technical and safety data sheets for this resin, the manual for the vacuum equipment, and a person trained to use it. A vacuum chamber can implode, so never convert an ordinary container. Do not use vacuum unless the resin maker allows it.

Back to chemistry ↑