There are three main families of analysis for precious metals in catalyst, industrial residue or ore: XRF (X-ray fluorescence), ICP-MS / ICP-OES (inductively coupled plasma) and fire assay (cupellation). Each does something different. Confusing them costs money in two directions: paying for expensive analysis that did not fit the case, or making commercial decisions on top of insufficient readings.
XRF — fast triage
X-ray fluorescence measures elemental composition by irradiating the sample with X-rays and detecting the fluorescence emitted by atoms. Done in seconds, non-destructive, no chemical preparation.
Good for:
- Code-identified automotive catalyst — fast read on the whole piece or the extracted ceramic.
- Triage of heterogeneous lots — separating rich from poor fraction before moving material.
- Same-day commercial decisions (price on arrival, pre-validation of an offer).
- Verifying the supplier's grade declaration.
Poor for:
- Powder, cake or sludge without volumetric preparation — surface XRF systematically underestimates.
- Detection in ppb. Typical XRF works at ppm; below that, precision drops.
- Highly heterogeneous fluorescence matrices (gradient ceramic, oxidised metallic support).
- Replacing a certified report in commercial operations sensitive to exact grade.
Cost: in-house XRF has nearly zero marginal cost per read. First tool to use.
ICP-MS — ppb precision
Inductively coupled plasma mass spectrometry. The sample is digested in acid, nebulised into a ~7,000 °C plasma, ionised, and the ions are separated by mass in a spectrometer. Typical detection in ppb.
Good for:
- Precise quantification of Pt, Pd, Rh, Ir, Os, Ru and other precious metals at low concentrations.
- Powder, cake, sludge or solution material, where XRF cannot deliver.
- Commercial decisions on large lots where 5% relative error is meaningful money.
- Operations with refiners or international buyers requiring a certified report.
Poor for:
- Decisions in minutes. Chemical preparation + read time + lab report: typically days to weeks.
- Triage of many small lots. Per-unit cost does not pay off.
- Refractory mineral matrices — ICP-MS reads what is in solution. If digestion does not dissolve everything, a fraction is lost.
Cost: varies by element, replicates and lab. Per sample, an order of magnitude above XRF. Worth it when the lot justifies.
ICP-OES — between the two
Same general technique as ICP-MS, but using optical emission spectrometry rather than mass. Typical detection in ppm — more sensitive than XRF, less than ICP-MS. Cheaper and more robust than ICP-MS for tricky matrices.
When it fits: matrix characterisation with several metals (Ni, Mo, V, Co, Cu, Fe besides PGM), where ppb does not change the decision and ppm is enough. Spent industrial catalyst frequently lands here.
Fire assay — the mineral reference
Cupellation: sample mixed with flux (litharge, carbonate, borax), fused in a high-temperature furnace, the precious metallic phase separated, cupelled to isolate the precious metal button. Destructive technique, old, expensive, still the gold standard for ore and concentrate.
When to use:
- Ore, concentrate and metal-bearing tailings in refractory mineral matrices.
- Operation with international buyer or refiner requiring fire assay as final proof.
- Large lots where the commercial operation pays for the assay without question.
When NOT to use: automotive catalyst (XRF + ICP-MS handle it), pharmaceutical filter (organic matrix complicates fusion), e-waste (mixed matrix needs another route).
How to decide in practice
- Always start with XRF. First sieve. Already eliminates lots not worth expensive assay.
- If material is identified ceramic and decision is in minutes: XRF may be enough.
- If lot is large, heterogeneous or non-ceramic: XRF triage, ICP-MS or ICP-OES confirm.
- If mineral ore/concentrate or buyer requires: fire assay as final proof.
- Always require documented chain of custody from collection to lab. Without it, the number does not defend a commercial decision.
The most common mistake
Setting sale price based on surface XRF without volumetric sampling on powder, cake or heterogeneous material. XRF reads what is exposed; if surface sample does not represent the lot, the number is the average of the shell, not the load.
Second most common: commissioning ICP-MS on a small lot whose result will not change decisions. Analysis cost becomes sunk cost. XRF would have been enough.
To define the right technique for your case, describe the material and objective in the submission form or WhatsApp. In some cases the answer is "XRF handles it, proceed", in others "worth coordinating ICP-MS". The wrong technique costs more than it looks.