XRF Gold Plating Thickness Measurement: Practical Limits and Use Cases

XRF can measure coating thickness, and it does so non-destructively in seconds. That capability is genuinely useful — and it is also routinely oversold. The technique has a defined working window, and outside that window it returns numbers that look valid and are not.

This guide sets out where the window begins and ends, what governs its edges, and which alternative method to reach for when a coating falls outside it. The intent is practical calibration of expectations: knowing when to trust a thickness reading is more valuable than knowing how to take one.

How XRF infers thickness at all

Composition analysis and thickness measurement use the same spectrum but ask different questions of it. Thickness measurement exploits a straightforward physical relationship: a coating both emits its own fluorescence and attenuates fluorescence coming from the substrate beneath it.

Two complementary signals result:

• Coating emission grows with thickness. A thicker gold layer contains more gold atoms in the analyzed volume, so gold intensity rises.
• Substrate emission shrinks with thickness. Nickel or copper fluorescence from below must pass through the gold to escape, and thicker gold absorbs more of it.

The ratio between these two moves monotonically with thickness, and the instrument solves for the thickness value that reproduces the observed ratio. Using both signals rather than one makes the calculation far more robust, because factors that scale both signals equally — geometry drift, minor intensity variation — largely cancel in the ratio.

Why the relationship is exponential rather than linear

Attenuation follows the Beer-Lambert law: intensity decays exponentially with path length through the absorbing layer. Thickness is therefore derived from a logarithm of an intensity ratio, not from a proportion.

This has an important practical consequence. Precision is not uniform across the thickness range. In the region where the exponential is steep, a small change in measured intensity corresponds to a small change in thickness, and precision is excellent. Where the curve flattens, the same intensity change corresponds to a large thickness change, and precision collapses. The flattening is what defines the upper limit of the technique.

The working window: roughly 0.05 to 5 microns

For gold on typical substrates, XRF thickness measurement performs reliably between about 0.05 µm and 5 µm. Both boundaries are physical rather than arbitrary.

The lower bound: too little signal

Below roughly 0.05 µm — 50 nanometres — the coating contains too few atoms to generate statistically meaningful fluorescence above background in a practical measurement time. The substrate signal is barely attenuated at all.

Very thin decorative flash plating, typically 0.05–0.1 µm, sits right at this boundary. XRF can usually confirm that some gold is present, but quantifying how much becomes unreliable. Extending the count time helps, following the 1/√N relationship, but the improvement is slow and the practical floor does not move far.

The upper bound: saturation

The upper limit is more important commercially because it is where confident wrong answers appear.

As coating thickness increases, substrate fluorescence weakens exponentially. Beyond a certain point, essentially no substrate signal escapes. The coating has reached saturation thickness — also called infinite thickness — and further increases produce no measurable change in the spectrum.

At that point the instrument sees exactly what it would see from solid bulk gold. It has no information distinguishing a 10 µm coating from a 50 µm coating from a solid gold item, because all three produce an identical spectrum.

What saturation thickness actually is

Saturation depends on the energy of the line being measured and the density of the material. For the substrate lines commonly used in gold plating measurement, practical saturation for a gold overlayer arrives in the region of 5–10 µm, depending on which substrate line is monitored.

Higher-energy substrate lines penetrate thicker gold and therefore extend the range. Monitoring a substrate’s K line rather than its L line, where the element permits, is the standard way to push the ceiling upward.

Why heavy gold plating defeats the technique

The commercially critical case is heavy gold plating and gold-filled material, and it sits squarely beyond the saturation ceiling.

• Standard decorative plating: 0.05–0.5 µm — well within range
• Hard gold for contacts and connectors: 0.5–2.5 µm — within range
• Heavy gold plating on jewelry: 2.5–10 µm — upper portion at or beyond the ceiling
• Gold filled, US standard: typically 1/20 by weight, often 20–50 µm — far beyond the ceiling

A gold-filled item with a 30 µm layer is, from XRF’s perspective, indistinguishable from solid karat gold. The instrument will report a composition consistent with solid gold, because the volume it can see genuinely is solid gold.

This is not an instrument defect and no amount of counting time or detector quality corrects it. The information required to distinguish the two cases never reaches the detector.

Multilayer stacks and the nickel barrier

Real coatings are rarely a single layer on a bare substrate.

A typical decorative stack is gold over nickel over brass. The nickel serves as a diffusion barrier preventing copper from migrating into the gold, and as a levelling layer improving appearance. Electronics finishes often use gold over palladium over nickel over copper.

Why the barrier layer complicates measurement

Thickness models must account for every layer between the surface and the point where signal originates. A two-layer model applied to a three-layer stack will produce a thickness value, and that value will be wrong — often substantially.

The nickel barrier is particularly consequential because nickel absorbs strongly in the energy region where copper fluoresces. If the software assumes gold directly on brass but nickel is actually present, the suppressed copper signal is misattributed to a thicker gold layer than exists. The reported gold thickness comes out high.

What multilayer measurement requires

Genuine multilayer capability requires the instrument to know the stack structure — which layers, in what order, of what materials — and to solve for multiple unknowns simultaneously. Each additional layer adds an unknown and degrades the precision of all the others.

Practical guidance: two layers are routine, three are feasible with a well-configured method and known materials, and four or more are specialist work. If you do not know what is underneath the gold, treat any thickness number with real suspicion.

What the numbers mean by application

Thickness specifications carry different weight depending on why the coating exists. A figure that is generous in one industry is a failure in another.

Decorative jewelry plating

Fashion jewelry commonly carries 0.05–0.25 µm. This is enough for appearance and essentially nothing else — such coatings wear through at contact points within months of regular wear. Sits comfortably inside the XRF window, so incoming inspection is straightforward.

Vermeil, by the US definition, requires at least 2.5 µm of gold over sterling silver. That figure is near the upper portion of the reliable XRF range, and the silver substrate complicates matters because silver’s own lines fall in a region that interacts with the measurement. Verifying vermeil compliance is a case where XRF gives a useful indication and coulometric stripping gives the defensible answer.

Electrical contacts and connectors

Hard gold on connector contacts typically runs 0.75–1.25 µm over a 1.25–5 µm nickel barrier. Both layers matter functionally — the gold prevents oxidation of the contact interface, the nickel prevents copper diffusion into the gold.

This is XRF’s strongest application. Both layers sit in range, the stack is known and constant, and the measurement is fast enough for line-side sampling. Nearly all plating shops in electronics use XRF for exactly this.

Semiconductor and high-reliability finishes

ENIG finishes on printed circuit boards specify gold in the 0.05–0.1 µm range over 3–6 µm of electroless nickel. The gold is deliberately thin, since its role is only to protect the nickel from oxidation before soldering.

The thin gold sits at the low end of the XRF window, and measurement requires longer counting for acceptable precision. The thicker nickel layer, by contrast, is measured easily.

Industrial wear coatings

Where gold is applied for wear or corrosion resistance rather than appearance, thicknesses of 5–25 µm are common. Most of this range is beyond XRF saturation, and the industry accordingly relies on coulometric methods and cross-sections for verification.

Building a defensible inspection routine

For a business receiving plated goods, the goal is a routine that catches problems at acceptable cost rather than exhaustive testing of everything.

Establish what the stack should be before testing

Ask the supplier for the specified layer structure, materials, and nominal thicknesses. Without this, any measurement is guesswork, since the model needs to know what it is solving for. A supplier unwilling to state the stack is itself informative.

Verify against a known reference regularly

Certified coating thickness standards exist for common combinations. Measuring one at the start of each session detects instrument drift before it contaminates a batch of results. This takes under a minute and is the single highest-value habit in coating inspection.

Sample by position, not at random

Because plating is thickest at edges and thinnest in recesses, random sampling produces scatter that reflects geometry rather than quality. Defining fixed measurement positions — the same points on every unit — makes results comparable across batches and across time.

Record the raw spectrum, not just the number

When a dispute arises months later, the composition table alone rarely settles it. The stored spectrum allows reanalysis with a corrected model if the assumed stack turns out to have been wrong. Instruments that permit spectrum archiving make this trivial and it costs nothing.

Define an escalation threshold in advance

Decide beforehand what result triggers destructive verification — a reading below specification by a defined margin, an unexpected substrate element, or unexplained spot-to-spot variation. Deciding in the moment, with a valuable lot waiting, tends to produce optimistic judgments.

Sources of error you can actually control

Assuming the coating sits inside the working window and the stack is correctly modelled, several practical factors still dominate real-world accuracy.

Surface roughness

Thickness models assume a flat, smooth interface. A rough surface presents a longer average path through the coating for emerging substrate radiation, biasing the result high. Heavily textured, brushed, or sandblasted finishes can shift readings by 10% or more.

Curvature

Curved surfaces such as ring shanks change the effective take-off angle across the measured spot. Where the surface curves away from the detector the path length increases, again biasing thickness high. Measuring on the flattest available area is the simple mitigation.

Contamination

Any layer on top of the coating — polish residue, lacquer, tarnish, skin oil, adhesive — attenuates the signal and adds apparent thickness. Cleaning with isopropyl alcohol before measurement is not fussiness; on a lacquered piece it can change the answer by a substantial fraction.

Coating composition assumptions

Models assume a known coating density. If the method assumes pure gold but the coating is actually a hardened gold alloy containing cobalt or nickel, density differs and the calculated thickness is systematically off. Hard gold plating for electrical contacts is commonly alloyed for wear resistance, so this is not an edge case.

Substrate composition assumptions

The same applies underneath. A method configured for brass applied to a bronze or nickel-silver substrate produces a wrong attenuation calculation and therefore a wrong thickness.

Choosing the right method for the thickness you have

XRF is one option among several, and each has a range where it is clearly the correct tool.

XRF

Non-destructive, fast, multi-element, no sample preparation. Best from 0.05 to about 5 µm. Fails silently above saturation. Ideal for incoming inspection, process control on plating lines, and rapid screening.

Coulometric stripping

Anodically dissolves the coating and measures the charge required. Destructive but only in a small defined area. Accurate over a wide range, including well beyond XRF saturation. Slower, requires electrolyte and technique, and is limited to one layer at a time. This is the standard reference method when XRF says “solid” and you need certainty.

Cross-sectional metallography

Cut, mount, polish, and measure the layer under a microscope. Fully destructive and slow, but it is the definitive answer, works at any thickness, and reveals structure that no indirect method can — porosity, diffusion zones, layer count. This is the arbitration method in a dispute.

Magnetic and eddy current gauges

Very fast and inexpensive but restricted to specific substrate and coating combinations, and they measure total non-magnetic thickness rather than identifying materials. Useful on a production line where the stack is known and constant, and unsuitable where composition is in question.

Weight difference

Weigh, strip the coating, weigh again. Simple and reliable for average thickness across an entire part, but destructive and gives no spatial resolution. Notably, this is the practical way to confirm a gold-filled ratio, since gold-filled is defined by weight fraction rather than by thickness.

A workable decision procedure

The following sequence resolves nearly every practical case without over-testing.

Step one: establish whether saturation is plausible

Run a standard XRF measurement. If a substrate signal is present, the coating is below saturation and a thickness measurement is meaningful. If no substrate signal appears at all, you are at or beyond saturation and the thickness number, if the software offers one, carries no information.

Step two: cross-check mass against volume

For jewelry, weigh the item and estimate its volume. Solid 18K gold has a density near 15.6 g/cm³; brass is around 8.5 g/cm³. A gold-plated brass item weighs substantially less than a solid gold item of the same dimensions. A significant density shortfall alongside a “solid gold” XRF result is the classic signature of heavy plating or gold-filled construction.

Step three: test multiple locations

Measure a high-wear edge, a flat face, and a clasp or joint. Plating is thinnest where wear is greatest, so a worn edge often reveals substrate that a pristine face conceals. Variation between spots is itself diagnostic — genuine solid material reads consistently.

Step four: escalate when value justifies it

If the item is high value and steps one to three leave ambiguity, move to coulometric stripping or, in a dispute, cross-sectional metallography. The cost of destructive testing on one piece is small against the cost of a wrong valuation on a lot.

The full counter-side version of this sequence is set out in our plated versus solid karat tutorial, and the retail application in the gold-coated jewelry verification guide.

Reading a thickness specification critically

When evaluating whether an instrument can do what a supplier claims, three questions cut through most marketing language.

What is the stated range, and for which coating and substrate pair?

A range quoted for gold on nickel does not transfer to gold on brass or gold on stainless. Saturation depends on both materials.

How many layers can the method solve simultaneously?

Single-layer capability is universal. Genuine two-layer solving is common. Three-layer requires a properly configured method and known materials. Vague claims of “multilayer” without a stated layer count usually mean two.

What precision is quoted, and at what thickness?

Precision varies dramatically across the range because the underlying relationship is exponential. A specification of ±2% is meaningless without stating the thickness at which it applies. Ask for the figure at the low end and the high end of the claimed range, not the midpoint.

Frequently asked questions

What is the thickest gold plating XRF can measure?

Practically around 5 µm for common substrate and line combinations, extending toward 10 µm when a higher-energy substrate line is available. Beyond that the substrate signal is fully attenuated and the coating is indistinguishable from bulk gold.

Can XRF tell gold-filled from solid gold?

Usually not by itself. Gold-filled layers are typically 20–50 µm, far beyond saturation, so the spectrum matches solid gold. Density cross-checks, high-wear-edge testing, or destructive confirmation are required.

Why does my thickness reading change between spots on the same item?

Plating is genuinely non-uniform. Electroplating deposits more heavily on edges and points and more thinly in recesses and interior surfaces. Variation is real information about the coating, not instrument error.

Does the substrate material affect the measurement?

Substantially. The substrate determines which lines are available for monitoring attenuation, how strongly it fluoresces, and where saturation falls. A method configured for one substrate applied to another produces systematic error.

How thin a coating can XRF detect?

Detection of presence is possible well below 0.05 µm with extended counting. Reliable quantification becomes difficult below that, because the coating contributes too little signal relative to background.

Is XRF thickness measurement destructive?

No. It leaves no mark and the item can be measured repeatedly. This is its principal advantage over coulometric and metallographic methods, both of which consume material.

Why does a lacquer coating change my result?

Lacquer and other organic coatings are largely invisible to XRF because light elements fluoresce too weakly to detect, but they still attenuate radiation passing through them. The attenuation is misattributed to a thicker metal layer, biasing the reading high.

Can XRF measure thickness on curved jewelry?

Yes, but with degraded accuracy. Curvature alters the effective take-off angle and lengthens the average escape path, biasing readings high. Measuring the flattest accessible area and treating curved-surface results as indicative rather than quantitative is the sound approach.

The essential limitation, stated plainly

XRF thickness measurement works by observing how much substrate signal a coating blocks. When the coating blocks all of it, the method has nothing left to observe.

Everything else — multilayer complexity, roughness, curvature, contamination — is a correctable error. Saturation is not correctable. It is an information limit, and the correct response is recognizing that you have hit it and switching methods, rather than trusting a number the physics cannot support.

For the underlying emission and absorption mechanisms, see our guide to the physics of XRF alloy composition. For how detector resolution affects which substrate lines can be monitored cleanly, see the detector comparison.

Two cases worth walking through

Abstract limits become clearer against concrete situations. Both of the following are ordinary occurrences rather than unusual edge cases.

A bracelet that reads as solid 18K but is not

A heavy link bracelet arrives at a buying counter. XRF returns 75.1% gold, 12.4% copper, 12.5% silver — a textbook 18K yellow composition, consistent across three spots on the flat outer faces.

The weight, however, is 42 g for a piece whose dimensions suggest closer to 68 g if solid. That gap is the entire story. A 30 µm gold-filled layer sits far beyond saturation, so every spot the operator tested was genuinely 18K gold — just 30 microns of it over a brass core.

What resolved it was measuring the clasp tongue, a high-wear component where the layer had worn through. Copper and zinc appeared immediately alongside a collapsed gold percentage. The instrument was never wrong; the sampling strategy was incomplete until the worn area was included.

A connector batch that fails incoming inspection incorrectly

A batch of gold-plated connectors is specified at 0.80 µm gold over 2.5 µm nickel on copper. Incoming XRF inspection reports gold at 1.15 µm, well above specification, and the receiving team flags the supplier for over-plating.

The supplier disputes it, and coulometric stripping on the same parts returns 0.79 µm — within specification. The discrepancy traced to the instrument’s method being configured as gold-on-copper, omitting the nickel layer entirely. Nickel absorbs strongly in the region where copper fluoresces, so the suppressed copper signal was misread as extra gold.

Correcting the method to a three-layer model brought XRF into agreement with the reference method. The lesson generalizes: a thickness number is only as good as the stack model behind it, and a wrong model produces a confident, repeatable, wrong answer.

Documentation that survives a dispute

Coating thickness results become contested more often than composition results, because money is usually attached to a specification tolerance. A record that holds up needs more than a number.

• The assumed stack — layers, materials and order used by the method. This is the first thing challenged and the most common source of genuine error.
• Measurement positions, described specifically enough to repeat. “Three spots” is not reproducible; “flat outer face at 12, 4 and 8 o’clock” is.
• The reference standard result from the same session, demonstrating the instrument was in control when the measurement was taken.
• Count time and repeat count, since precision claims depend on both.
• The raw spectrum, which permits reanalysis under a corrected model without recalling the parts.

The spectrum matters most. When a stack assumption turns out to be wrong months later, archived spectra let you recompute; a stored number cannot be repaired.