XRF analysis of gold and precious metals: method, standards and limits

An assay office that has to declare the fineness of a gold article takes a fragment, melts it with lead in a cupel, dissolves the silver in nitric acid and weighs the gold that remains. That is the cupellation of ISO 11426:2021, and in Italy it is also the official method for gold. A metal analyzer based on X-ray fluorescence asks the same article the same question without touching it and answers in a few seconds, but it only interrogates the first few micrometers of metal. The two answers agree only when the alloy is homogeneous, the surface represents the core of the piece and the calibration suits that alloy family. The method of XRF analysis on precious metals comes down to knowing when those conditions hold and noticing in time when they don't.

Which lines the spectrometer reads on a precious alloy

The principle of X-ray fluorescence is the same as in any XRF analyzer for metals, and the guide to choosing a portable XRF analyzer sums it up. With precious metals, what matters is which lines you work on. Gold's K edge sits at 80.7 keV in the NIST tables, and exciting the K lines would take photons above that threshold. Precious-metal spectrometers therefore read gold on its L lines, Lα at 9.71 keV and Lβ at 11.44 keV according to the X-Ray Data Booklet from Lawrence Berkeley National Laboratory. Silver, palladium and rhodium are read on their K lines, between 20.2 and 22.2 keV, while nickel, copper and zinc fall between 7.5 and 8.6 keV.

Trouble starts where lines crowd together. Platinum's Lα falls at 9.44 keV, 270 eV below gold's, and zinc's Kβ at 9.57 keV slips in between the two; just below sits iridium's Lα at 9.18 keV. On platinum or palladium white gold, or on a zinc-rich yellow gold, the software has to separate partly overlapping peaks. It does better the narrower the detector's energy response, and this is where a silicon drift detector (SDD) earns its premium over a PIN diode. Tungsten is the extreme case, with its Lβ1 line at 9.67 keV is only 41 eV from gold's Lα.

How many micrometers the signal comes from

The second condition is depth. Fluorescence photons born inside the metal have to get out, and gold absorbs them strongly. Using NIST attenuation coefficients, a density of 19.3 g/cm³ and normal incidence, the intensity of gold's Lα drops to 1/e, just over a third, within about 4 micrometers of pure gold. Lβ at 11.44 keV still lies below gold's L3 edge (11.92 keV) and comes from about 6 micrometers; silver's Kα at 22.16 keV from about 9. In 750 alloys the depths grow a little, because copper and silver absorb less than gold, but they stay on the order of ten micrometers.

On a homogeneous piece this limit doesn't matter. It matters when surface and core differ, and the cases are well known. A few micrometers of electroplated gold are enough to switch off the signal from the metal underneath. Copper Kα at 8.05 keV, crossing 5 micrometers of gold, reaches the detector at about a seventh of its strength, and through 10 micrometers at less than 2%. The incoming beam, meanwhile, also loses intensity before it reaches the core. Pickling and finishing can deplete the surface of copper and silver and leave it richer in gold than the bulk, the surface enrichment known as depletion gilding. White gold is often rhodium-plated, and the rhodium in the spectrum belongs to the coating. Solder joints on chains and rings, finally, often have an alloy of their own, usually of lower fineness.

Coatings have a standard of their own. ISO 3497:2000, reconfirmed in 2022, covers coating-thickness measurement by X-ray spectrometry. The instrument determines mass per unit area and derives thickness from density, for up to three layers. It works when the layer structure is known in advance, so it is better at checking a declared gilding than at uncovering a hidden one. The sequence a buyer follows at the counter to expose a plated item, from readings at several points to filing, is laid out step by step in the article on verifying gold fineness at a gold buyer's counter.

From counts to fineness: calibration and measuring conditions

A spectrum is made of counts, and fineness comes from a model that turns them into concentrations. The model can start from the fundamental parameters of X-ray interaction with matter or from an empirical calibration on alloys of certified composition. Either way it has to correct for matrix effects. In a gold alloy, copper radiation is absorbed by the gold itself, while silver's K lines excite secondary fluorescence in gold and copper. The same copper intensity therefore means different concentrations in a 375 and in a 750. The piece itself plays a part too, as Marco Actis Grande documents in his paper on XRF as a fineness analysis tool at the 2016 Santa Fe Symposium. Geometry, surface finish, size and alloy all change the reading, and a fixed measuring protocol is needed to avoid producing false values.

Two practical rules follow from this. The first concerns calibration, which has to be built or checked on samples from the same alloy family and fineness range you will measure. A curve set up on yellow gold guarantees nothing on palladium white gold. The second concerns statistical precision, which improves with the number of counts collected and therefore with measuring time. At high fineness, where a few parts per thousand matter, you need longer times and reference samples close to the expected value. The standard that now describes this use, ISO 23345:2021, draws its boundaries precisely. Its non-destructive ED-XRF method serves to verify, that is to confirm, the declared fineness of finished and semi-finished jewelry considered homogeneous; it excludes any coated item and rules out wavelength-dispersive (WD-XRF) instruments.

Cupellation, gravimetry, ICP-OES: who sets the reference fineness

The fineness that counts is set by the destructive methods of technical committee ISO/TC 174, one per metal. For gold it is ISO 11426:2021, which applies cupellation to material considered homogeneous with a gold content preferably between 100 and 999.5‰. The standard provides variants for samples rich in base metals, platinum, palladium or silver. It also has a limit that concerns XRF users directly, because it is not compatible with iridium, rhodium and ruthenium above trace levels, meaning more than 0.25‰ combined. For Christopher Corti, who reviewed gold assay methods at the 2001 Santa Fe Symposium, it is the reference method by which the accuracy of every other one is judged. ISO 11426 is also the recommended method for alloys whose fineness range is set by ISO 9202:2026.

For the other metals the logic is the same and only the chemistry changes. Silver is determined by potentiometry under ISO 11427:2024, platinum by gravimetry under ISO 11210:2023 and palladium, also by gravimetry, under ISO 11490:2023. Above 999‰ the fineness is obtained by difference, measuring the impurities by ICP-OES and subtracting them, with ISO 15093:2020 for gold, platinum and palladium and ISO 15096:2020 for silver.

Read this way, the table shows where XRF fits. If the number has to end up on a hallmark, a certificate or in a dispute, then you need one of the destructive methods, run by a laboratory. If instead the question is whether a homogeneous article has the fineness it declares, ISO 23345 describes exactly that check. The spectrometer runs it in seconds on every piece rather than on a sample taken from one. There is also a less visible use. A spectrum taken before assay tells the laboratory whether palladium, platinum or iridium are present, which change the cupellation procedure or make it unusable.

What Italian law requires on fineness

In Italy the framework is set by Legislative Decree 251 of 22 May 1999. Article 3 requires fineness to be expressed in parts per thousand and guaranteed on melting for every part of the article. The legal finenesses are 750, 585 and 375 for gold, 925 and 800 for silver, 950, 900 and 850 for platinum, 950 and 500 for palladium. Any fineness above the highest one is allowed. Negative tolerances are ruled out, apart from a few listed exceptions for platinum, palladium and lost-wax cast gold. The same article defers to the implementing regulation, Presidential Decree 150 of 30 May 2002, for the official methods of analysis and the maximum permissible error. Commercial scales translate as follows:

The official methods were updated recently. The ministerial decree of 19 March 2025, published in Italy's Official Gazette no. 95 of 24 April 2025, renewed the regulation's technical references. For gold the method remains cupellation under UNI EN ISO 11426, while for platinum and palladium an ICP-OES method with internal standard now sits alongside gravimetry. Oversight of fineness is entrusted to the Chambers of Commerce, which sample articles on the market and send them to authorized laboratories for assay.

On international hallmarking, Italy joined the Vienna Convention on the control and marking of articles of precious metals through Law 55 of 15 May 2023. It has been listed among the Contracting States since 15 December 2023. The resulting Common Control Mark (CCM) is voluntary. In Italy it is applied by the assay offices of the Chambers of Commerce of Alessandria-Asti, Arezzo-Siena and Vicenza, as the Ministry of Enterprises and Made in Italy explains. The wholesale market follows rules of its own, and to enter the LBMA Good Delivery system a gold bar must be at least 995.0‰ fine and a silver bar at least 999.0‰. The fineness is stamped on the bar to four significant figures for gold and three for silver.

Where the XRF spectrometer works well, and where it stops

Put physics, calibration and standards together and the scope of XRF on precious metals draws itself. It works well where the alloy is homogeneous and the surface represents it. That covers incoming checks on grain, bars and semi-finished stock at a refinery, alloy verification in jewelry manufacturing and sorting recycled material by fineness. Faced with plated or rhodium-plated items, articles with extensive soldering and pieces smaller than the measuring spot, it stops or asks for an extra check. At a gold buyer's counter these cases are common, and that is where the procedure matters as much as the instrument. How to set it up is explained point by point in the counter procedure for gold buyers and bullion dealers.

Two variables weigh most on the choice of instrument: which metals you need to separate and how small the area to be measured is. The first decides detector and calibrations, the second the collimation and the aiming camera; format and the overall cost of an XRF analyzer follow from there. Configurations are collected on the page on XRF and PMI analyzers for material analysis and on the one devoted to the range of XRF analyzers we distribute. If your case falls in the range where XRF works well, it makes sense to start from a test on your own pieces, checked against a known reference fineness.

What to send us for an assessment

A few precise details are enough for us to propose a sensible configuration and test:

• metals and alloys to measure (yellow, white or rose gold, silver, platinum, palladium) and the most frequent finenesses;
• type of items: finished jewelry, semi-finished stock, grain, bars, recycled material, with the size of the smallest area to measure;
• coatings to recognize or measure in thickness, such as rhodium plating and gilding;
• maximum acceptable deviation from the declared fineness and the assay laboratory to compare results with;
• number of readings per day and place of use: counter, laboratory, production floor;
• samples of known fineness or assay certificates you already have.

You can send them by email to info@dipitech.com or on WhatsApp, attaching a few photos of the items you measure most often.