XRF testing: what it measures, what it cannot prove, and how a reading becomes a fineness

01

What an XRF analyser actually measures

A seller's real question about testing is whether the number can be trusted, and the answer starts with how the number is made. An XRF analyser sends a narrow beam of X-rays at a small spot on the item. The atoms in that spot answer with X-rays of their own, at energies particular to each element, and the analyser counts them and works out the proportion of each element in the volume the beam reached. For a gold alloy the result is a list such as gold 58.6 per cent, silver 8.1, copper 30.9, zinc 2.4, which the operator reads as a fineness.

The test takes seconds, leaves no mark and needs nothing cut from the piece. Those are real advantages over an acid test, which scratches, and fire assay, which destroys a sample. They are also why its limits need saying plainly, because a fast, clean test invites more confidence than it can carry on its own.

Penetration depth

How far into the metal the analyser gets a usable signal from. For a gold alloy it is roughly the outer fifteen to twenty microns, about a fiftieth of a millimetre. Anything deeper than that is invisible to the test.

Spot size

The area the beam covers: as small as 0.3 mm across on a benchtop instrument, and several millimetres on a handheld one. A reading describes that patch, not the whole item.

Reading

The elemental percentages reported for one spot. Taking several readings on different parts of a piece is normal practice, not a sign that something is wrong.

Interpreted fineness

The parts-per-thousand figure finally adopted for the item once the readings have been reconciled with the hallmark and the other checks.

02

What XRF cannot prove on its own

Because the signal comes from the surface, any item whose outside differs from its inside can give a reading that is true of the surface and wrong about the piece. No test that sees a fiftieth of a millimetre into an object can be called certain about what lies beneath.

• Plating. A thin coat, a micron or two over brass, usually lets the copper and zinc underneath show through in the reading, because the beam reaches them. A heavier coat does not: once the gold is deeper than the analyser sees at that spot, the reading describes the coat and is silent about the base metal beneath it.
• Gold-filled and rolled gold. A bonded sheet of carat gold over brass is thicker than plating and can read as solid carat gold across the whole surface. Weight, density and a look at the cut edges are needed.
• Solder. Joints and repairs are made with lower-fineness solder. A reading on a joint under-reads the item; a reading away from the joint does not reveal that a large repair is there.
• Hollow construction. A hollow 9ct bangle or rope chain reads exactly like a solid one. XRF says nothing about how much metal there is, only what the surface is made of.
• Internal fillings. Weighted silver, pitch-filled handles and cement under stones are outside the beam altogether.
• Surface enrichment. Some manufacturing and finishing processes leave the surface richer in gold than the body of the alloy, so the surface reads a little high.
• Geometry and size. Curved, very small or very thin pieces scatter or lose signal, and the readings drift away from the true composition.
• Contamination. Polish residue, skin oils, dirt and adhesive all distort a reading. Cleaning the spot alters the piece, so it is done only with your written consent (clause 4.2 of the terms).

03

How a reading is turned into a fineness

A reading is a percentage; a fineness is a standard. The two are reconciled, not equated. If a ring hallmarked 585 reads 58.4, 58.7 and 58.5 per cent gold at three spots away from the shank joint, the interpreted fineness is 585, because the readings agree with the mark within the analyser's normal variation and there is no reason to depart from the standard.

If an unmarked ring reads 58.4 per cent, the same fineness is adopted, but the item list says it rests on the test alone. If a ring marked 750 reads 58.4 per cent everywhere, the mark is treated as unreliable and the tested figure is used, with the discrepancy written down.

1. Read the hallmark firstThe sponsor's mark, fineness mark and assay office mark are recorded before any test, so the test is checking a claim rather than making one.
2. Weigh on a calibrated balanceGross weight is recorded to two decimal places. A piece that is lighter than its size suggests is a candidate for hollow or filled construction.
3. Magnet checkGold and silver alloys are not drawn to a magnet. A pull means steel or nickel somewhere: a clasp spring, a plated base or a loaded interior.
4. XRF at more than one spot, where an analyser is availableReadings are taken on a clean area of the main body and, where present, at joints, clasps and any area that looks different in colour or wear. If a spot has to be cleaned first, that is asked for in writing.
5. ReconcileIf mark, weight, magnet and readings agree, the marked or read fineness is adopted. If they do not, the disagreement is written down and a further test is proposed.
6. Escalate only with your permissionDensity testing is non-destructive and may be done as a matter of course. An acid test leaves a small mark and is never done without your written agreement; you can have the piece back untested instead. Fire assay would settle what the rest cannot, and there is no arrangement with a laboratory, so it is not something we can put to you.

04

The tests used alongside XRF and what each adds

Complementary tests and the question each one answers

Scroll sideways for more →

Density deserves a word, because it is the non-destructive check that catches most of what XRF misses. We weigh the piece in air and again in water; the two figures give a density that plating and brass cores cannot fake. Pure gold is about 19.3 g per cubic centimetre, a typical 9ct alloy around 11, and sterling silver about 10.3 to 10.4. A piece whose surface reads 18ct but whose density comes out near brass is not 18ct through.

Density has its own blind spot. A hollow rope chain or a bangle with a sealed air space traps air and reads far too light, and a set stone throws the figure the other way. On those pieces we lean on weight against size and on the marks instead, which is one more reason no single test is ever enough.

05

Three readings on one ring, and what they are read against

Most of this guide is about disagreement, so here is what agreement looks like, on an illustrative ring rather than a real case. Three readings never land on the same decimal. The question is only whether they scatter within the range the alloy would be expected to give, and whether they sit with the mark.

An illustrative hallmarked 585 ring, read at three spots on a clean surface

Scroll sideways for more →

The mark, three readings, the magnet and the balance all point the same way, so 585 is the fineness adopted, and the record says it rests on a hallmark confirmed by three readings rather than on either alone. Nothing more is done to the ring.

Now change one number. Had the shank joint read 45 per cent against the 58.4 on the body, the joint would be listed on its own as solder at its tested fineness, rather than the whole ring being written down to the lowest reading it gave. That is the practical value of recording where each reading came from: it keeps one small piece of lower-grade metal from dragging down the description of everything it touches.

And had all three readings come in near 45 per cent on a ring marked 750, the mark would be treated as unreliable, the disagreement written down, and a density check proposed before anything else happened. Nothing on this site turns a reading into a number you could work a price from; the readings decide what the item is, and what the lot is worth is a judgement made about the whole of it afterwards.

06

How test results appear on your item list

Every item Assaypost receives gets its own line on the item list, and the test evidence sits on that line rather than in a summary at the bottom. Clause 5.1 of the terms sets out what each line carries: the description, the recorded weight, any weight left out and why, the tests carried out, any test you asked us not to carry out, the fineness concluded and anything the tests could not establish. Where an analyser was used, each reading is listed with the spot it came from; until one is named, the list simply says which tests were used on each piece. One written offer for the whole lot follows the list, as one amount with nothing taken off it afterwards (clause 5.3).

07

Seller risks and exceptions

• A piece that reads as carat gold at the surface but is plated or filled may only be found out at the density or acid stage. Where that happens the item list says so, and you can decline the offer and have everything back, or ask for that piece back and get a fresh offer for the rest (clause 6.3 of the terms).
• Antique and hand-made pieces are often built from more than one alloy, with a different fineness in the shank, the setting and the decorative work. They are listed in parts, each at its tested fineness.
• Very small items, such as single earring backs or thin wire, give unstable readings. They are usually grouped and recorded at the lowest consistent reading, and the list says so.
• White gold is nearly always rhodium plated to make it whiter, and rhodium is itself a precious metal. The layer is thin and usually wears through at the shank, so a reading is taken on a worn spot; on a heavily re-plated piece the rhodium is noted on the item list as the reason a reading looked odd.
• Dental gold varies from one alloy to the next and often contains platinum-group metals; coloured gold alloys and some Asian and Middle Eastern high-carat pieces contain elements that shift readings too. These are interpreted with the alloy type in mind and the reasoning is recorded. The dental gold guide covers the first of these.
• No acid test is carried out unless you agree to it, and fire assay is not available here, so it is never proposed. Saying no is not held against you; the piece is either described on the non-destructive evidence with its limits stated, or sent back.

08

When not to use the scrap route

The testing and XRF page sets out the methods and what each one can settle, and the how it works page shows where testing sits between your photographs and your payment.