Handheld XRF has made positive material identification quick enough that it now happens routinely — pull the trigger, wait a few seconds, read the grade off the screen. That convenience has produced a widespread and expensive misunderstanding about what the instrument actually measured.
This article is mostly about what XRF cannot do, because that is the part that causes problems, and because we would rather tell you before you have a certificate that says something it should not.
How it works, briefly
X-ray fluorescence directs X-rays at a sample. That energy excites electrons in the atoms of the material, ejecting them from inner shells. Electrons from outer shells drop in to fill the vacancy, releasing energy as a secondary — fluorescent — X-ray at a wavelength characteristic of that element. The detector measures the energies coming back and the quantity at each, giving elemental composition.
It is fast, non-destructive, needs almost no preparation beyond a clean surface, and works on components in situ. For confirming that a pipe spool is the alloy the drawing says, it is excellent.
What it cannot see
The technique depends on an element producing a detectable characteristic X-ray. Light elements produce low-energy fluorescence that is readily absorbed — by the sample itself, by air, and by the detector window — before it can be counted.
In practice, handheld XRF cannot measure carbon. It also struggles with or cannot detect boron, beryllium, lithium, nitrogen and, on many instruments, sulphur and phosphorus at the levels that matter.
Carbon is the one that causes commercial damage, and here is exactly why.
The 304 / 304L problem
304 and 304L stainless steel are chemically almost identical in everything XRF can detect. Same chromium range, same nickel range, same iron balance. The distinction between them is carbon content — the L designating low carbon, typically capped around 0.03% against roughly 0.08% for standard 304.
Since XRF cannot measure carbon, it cannot distinguish 304 from 304L. The same is true of 316 and 316L, 321 and its variants, and every other L-grade pair.
An XRF instrument presented with either will match the elemental fingerprint and report a result. Some instruments display "304/304L" honestly. Others display whichever sits first in their library, and an operator who does not understand the limitation writes "304" on the report. That report is now making a claim the instrument did not support.
This is why our PMI reports state the pair — 304/304L — rather than picking one. It is less satisfying to read and it is what the measurement actually justifies.
Why the carbon grade matters
If it were a paperwork distinction nobody would care. It is not.
Carbon content governs susceptibility to sensitisation. When austenitic stainless steel is held in roughly the 450–850°C range — which happens in the heat-affected zone every time it is welded — carbon migrates to the grain boundaries and combines with chromium to form chromium carbides. That locally strips chromium out of the surrounding metal, and chromium is what provides the passive layer that makes stainless steel corrosion resistant.
The result is a grain boundary network depleted in chromium and preferentially attacked by corrosion. This is intergranular corrosion, and in the presence of chlorides and stress it contributes to stress corrosion cracking.
L grades exist precisely to limit this. Less carbon means less carbide formation means less sensitisation in the weld HAZ.
So a component installed in a chloride-bearing or elevated-temperature service, specified as 316L, and actually made of 316, is not a documentation discrepancy. It is a component with a materially different service life in an application where the difference was the reason for the specification.
When you genuinely need the carbon number
Not always. XRF is entirely adequate for most alloy verification. You need carbon measured when:
- The specification calls for an L grade and the service environment is the reason.
- The component will be welded and then run in a corrosive or elevated-temperature service.
- You are investigating a corrosion failure and need to know whether sensitisation was a factor.
- A code, client specification or regulator requires full chemical composition rather than elemental verification.
- You are distinguishing between carbon steel grades, where carbon content is the primary distinguishing property.
What to use instead
Optical emission spectroscopy is the practical answer. OES vaporises a small amount of the surface with an electrical arc or spark and analyses the light emitted. It measures carbon, and portable OES instruments bring that capability to site.
The trade-off is that OES is slightly destructive — it leaves a small burn mark, a few millimetres across — and needs more surface preparation. On a finished, coated or cosmetically critical surface that matters. On a pipe spool it does not.
Laboratory analysis of a removed sample gives the most complete and accurate result, including trace elements, where the component can be sampled and the timescale allows.
The sensible pattern on most jobs is XRF for broad screening across many components, with OES or laboratory analysis on the subset where the carbon grade actually carries consequences.
Where XRF is genuinely definitive
Having spent this long on the limitation, it is worth being equally clear about where XRF settles a question outright — because that covers most of the work.
Anywhere the distinguishing elements are heavy enough to detect, XRF is decisive. Separating 304 from 316 is trivial: 316 contains roughly 2–3% molybdenum and 304 contains essentially none, and molybdenum fluoresces strongly. Distinguishing carbon steel from any stainless is immediate on chromium alone. Identifying nickel alloys is straightforward, since their signatures are dominated by nickel, chromium, molybdenum and iron in distinctive proportions.
Duplex versus austenitic is reliable on chromium and molybdenum content. Titanium, copper alloys, brasses and bronzes all read clearly.
The classic industrial application — confirming that a spool marked as one alloy is not quietly another after a shutdown, or verifying that the right material went into a pressure boundary — is exactly what XRF is best at. Material mix-ups are usually between different alloy families rather than between a grade and its L variant, and XRF catches those instantly and at scale.
So the guidance is not "distrust XRF". It is: use it for what it measures, screen everything with it, and escalate the specific subset where carbon carries consequence.
Other things worth knowing about XRF
It measures the surface. Penetration is shallow — microns. Coatings, plating, oxide scale, weld spatter and grinding smear all read as the material. Surface preparation is not optional, and a reading taken through paint is a reading of the paint.
Weld metal is not parent metal. Filler is often deliberately over-alloyed. Reading across a weld gives a composition that matches neither the parent nor the filler specification. Take readings on parent material, clear of the weld, unless you specifically intend to analyse the weld metal.
Reading time affects accuracy. Short readings give a fast approximate result; longer counts improve precision, particularly for lighter detectable elements. An operator under time pressure taking one-second readings is producing screening data, not verification data — a distinction that should appear on the report.
Calibration and verification matter. The instrument should be checked against certified reference materials representative of the alloys being tested, at a defined frequency, with those checks recorded. A PMI report without evidence of instrument verification is an assertion.
It is a radiation-emitting device. Handheld XRF units are subject to IRR17 in the UK, with the accompanying requirements for risk assessment, appointed persons, and operator training. Not an onerous regime for these instruments, but a real one.
What a PMI report should say
A report you can rely on states the instrument and its calibration status, the surface preparation, the reading duration or mode, the measured elemental composition with the actual figures rather than only a grade name, the reference standard the result was matched against, and — explicitly — that carbon was not measured where XRF was the method.
Where the result cannot separate an L grade from its standard counterpart, it should say so, and report the pair.
That last item is the one that distinguishes a useful report from a reassuring one. We would rather hand you a result with a stated limitation than a clean-looking certificate that quietly overstates what the instrument did — because the second kind only becomes a problem years later, in a corrosion investigation, when somebody asks how the grade was determined.
We run PMI by both XRF and OES, on site and in our Cumbria laboratory. If your scope involves L grades in corrosive service, talk to us about which components genuinely need the carbon number and which do not — it is usually fewer than people expect, and knowing which is which is most of the saving.
