The Gold Standard of Archaeology | Portable XRF, Calibration Standards and Matrix Effects

1. Begin with the decision the measurement must support

Back to contents

Decide whether the goal is qualitative screening, grouping, compositional comparison or defensible concentration. A field survey looking for elevated lead does not need the same calibration as provenance work that separates visually similar obsidian sources. Name the material, elements, expected range and scale of inference. The instrument’s factory mode may be useful for triage but should not silently become a universal archaeological calibration. Analytical design begins with what difference must be detected and what error would change the interpretation.

2. Understand what reaches the detector

Back to contents

The source irradiates a small area; emitted X-rays escape from element- and matrix-dependent depths before reaching the detector. Peak energies help identify elements, while peak intensities also depend on concentration, absorption, enhancement, tube settings, filters, geometry and counting time. Light elements are especially difficult in air and through rough or contaminated surfaces. The displayed number is therefore the output of a physical and software model, not a direct reading of the object’s entire composition.

3. Treat the sample matrix as an active participant

Back to contents

A matrix is the surrounding material in which target elements occur. Silicate glass, porous ceramic, carbonate-rich soil and metal alloy absorb and enhance X-rays differently. Moisture, grain size, mineralogy, density and surface roughness can alter results even at the same true concentration. Calibrating with metal standards and applying that line to ceramic paste is not a harmless shortcut. Use reference materials that resemble the archaeological matrix and concentration range, or restrict claims to reproducible relative signals.

4. Standardise surface, geometry and acquisition

Back to contents

Keep the beam spot, instrument-to-sample distance, angle, measurement time, voltage, current and filter settings consistent. Document whether surfaces were cleaned, abraded, flattened, powdered or measured through a bag, coating or corrosion layer. Destructive preparation may improve representativeness but conflict with conservation. For intact objects, take multiple readings over documented locations and interpret variation as information about heterogeneity and surface history rather than averaging it away automatically.

5. Calibrate and validate with different evidence

Back to contents

Calibration fits the relationship between measured response and reference values; validation tests that relationship on standards not used to fit it. Use matrix-matched certified reference materials where available, supplemented by well-characterised secondary standards when necessary. Cover the intended concentration range without extrapolating beyond it. Report the model, transformations, residuals and element-specific performance. A high calibration correlation can coexist with biased predictions, so independent validation and check standards matter more than a visually straight line alone.

6. Worked example: strontium in a ceramic sherd

Back to contents

A factory soil mode reports 180 parts per million strontium on a ceramic sherd. A matrix-matched calibration built from ceramic standards converts the same spectral response to 145 ppm; an independent ceramic reference with a certified value of 150 ppm is recovered at 147 ppm. The closer validation supports the ceramic calibration for that range. It does not prove the whole sherd is 145 ppm: three surface spots range from 138 to 159 ppm, so the analyst reports within-object heterogeneity and the measured locations.

7. Monitor drift, blanks and repeatability

Back to contents

Measure a blank to identify contamination and instrument background, a stable check standard at intervals to detect drift, and replicates to estimate short-term precision. Randomise sample order or bracket batches so time is not confounded with archaeological group. Track battery state, temperature and software or calibration version when relevant. Precision describes repeatability, not closeness to truth: an instrument can return nearly identical values that are consistently biased because the calibration or matrix is wrong.

8. Changed-context transfer check: obsidian and sediment

Back to contents

Move from homogeneous obsidian to damp occupation sediment. The same instrument and element labels do not preserve the same evidence chain. Obsidian sourcing may use carefully prepared surfaces and source-specific reference collections; sediment moisture, grain size and vertical mixing can dominate field measurements. The transferable method is to restate the target, match standards and validate performance in the new matrix. Calibration coefficients should not travel merely because the analyser does.

9. Separate compositional groups from provenance claims

Back to contents

Elemental similarity can support groups, but provenance additionally requires representative source samples, geological plausibility, manufacturing and exchange context, and attention to recycling or mixing. Multivariate plots can separate clusters because of surface treatment, instrument drift or batch order rather than origin. Include quality-control materials in ordinations, evaluate classification error and test unknowns independently. Say “compositionally consistent with” when the data do not uniquely identify a source.

10. Know the limits of surface and detection

Back to contents

Weathering, patina, burial deposits, conservation materials and corrosion can make the surface unlike the bulk. Thin objects or irregular cavities may violate thickness and geometry assumptions. Peak overlaps and values near detection limits require spectral inspection, not blind acceptance of software flags. Some elements are poorly measured under field conditions. Non-destructive does not mean non-invasive in interpretation: repeated radiation is usually physically minor, but an overconfident classification can still damage the archaeological record.

11. Audit spectra, standards and decisions together

Back to contents

Retain raw spectra, instrument model and serial number, firmware and software, settings, beam spot, preparation, measurement coordinates, acquisition order, blanks, replicates, check standards and calibration files. Publish standard reference values and validation residuals. Reanalyse spectra after excluding suspect surfaces, changing matrix correction or using alternative grouping rules. An auditable archive allows later researchers to distinguish a new archaeological interpretation from a changed calibration or software routine.

12. The gold standard reports a calibrated surface measurement

Back to contents

Portable XRF becomes strong archaeology when speed does not outrun measurement science. The gold standard defines the analytical purpose, controls geometry and preparation, uses matrix-matched standards, validates independently and keeps raw spectra attached to every reported concentration. It distinguishes repeatability from accuracy and chemical grouping from provenance. The final claim should say which surface, matrix, element range and calibration support it—and where those supports end.