What can PMI testing with XRF verify on a metal part, and what can it miss?

What does a handheld XRF analyzer actually measure?

It measures the elements present in a thin surface layer of the part by exciting the atoms with X-rays and counting the characteristic X-rays each element sends back. It reports a chemistry, then matches that chemistry against a grade library; it never measures strength, hardness or structure.

Every element fluoresces at its own energies, so the analyzer can sort the returning signal by element and estimate how much of each is present. That works well for the heavier alloying elements that define most grades: chromium, nickel, molybdenum, manganese, copper, zinc, lead, titanium, niobium and tungsten. It works less well as elements get lighter, because their fluorescence is weak and is absorbed by air and by the detector window before it is counted. Current handheld units can report magnesium, aluminum, silicon, phosphorus and sulfur, but with poorer precision and usually only with longer test times; older units may need a helium purge to see them at all. No handheld XRF measures carbon, nitrogen or boron.

Two consequences follow. First, the reading describes the surface the window touches, not the core. A zinc-plated bracket reads as zinc, a powder-coated housing reads as the coating, and a part with heavy scale or contamination gives a distorted answer. Test bare stock, a masked spot, or a small area cleaned to bright metal. Second, the grade name on the screen is a library match, not a conformance decision. When two grades in the library differ only in an element the analyzer cannot see, the screen may name either, or both. Read the element values against the specification limits, not the grade label.

ASTM E1476, the guide for metals identification, grade verification and sorting, frames this well: some methods measure composition directly, while eddy current and thermoelectric methods respond to properties affected by both chemistry and processing and give only indirect information. Choosing a method starts with knowing which of those two questions you are asking.

Why can’t XRF confirm carbon content or tell 304L from 304?

Carbon is too light for X-ray fluorescence to detect in a handheld instrument, so any grade pair defined by carbon is invisible to it. Confirming carbon takes spark optical emission spectrometry (OES) or a combustion analysis in a lab.

The L grades of stainless exist for one reason: lower carbon limits sensitization during welding. If the drawing says 316L because the part is welded or sits in a corrosive service, a passing XRF reading proves only that the part is a 316-family alloy. The carbon question is still open. The same applies to 1018 against 1045, where the difference in carbon is the difference between a soft weldable steel and one that hardens, and to 52100 bearing steel, whose chromium XRF can see but whose high carbon content it cannot.

Spark OES works by striking an arc on a prepared spot and reading the light emitted, and with an argon-shielded spark it measures carbon, sulfur and phosphorus along with the metallic elements. The laboratory method for carbon and low-alloy steel is ASTM E415, and austenitic stainless is covered by ASTM E1086. Mobile OES units bring the same principle to the shop floor, at the cost of a small burn mark that must sit on a non-cosmetic, non-sealing face. For the most defensible carbon and sulfur numbers, a lab burns a chip sample and measures the combustion gases per ASTM E1019, which is destructive but unambiguous. Some handheld laser (LIBS) analyzers also report carbon; if you accept one, require that its calibration covers the low-carbon range your grade needs, and that the report states the result against the specification limit.

The comparison of the L grades themselves is in 316 vs 316L stainless. What matters here is the instruction: when the grade difference is carbon, the PMI method on your purchase order must be one that measures carbon.

How should you write a PMI requirement on a drawing or purchase order?

Name the method, the elements that must be reported, the sample size, the stage at which the test happens and what the report must contain. “PMI required” on its own invites a handheld reading with a grade label and nothing else.

Useful wording, adapted to the grade on the drawing, looks like this:

• MATERIAL VERIFICATION: PMI BY XRF OR OES PER ASTM E1476 GUIDANCE, EACH HEAT OF BAR STOCK BEFORE MACHINING. REPORT CR, NI, MO, MN WITH HEAT NUMBER AND INSTRUMENT ID.
• 316L ONLY: CARBON TO SPECIFICATION LIMIT, VERIFIED BY SPARK OES OR COMBUSTION ANALYSIS. HANDHELD XRF NOT ACCEPTED FOR CARBON.
• 6061-T6: TEMPER VERIFIED BY HARDNESS AND EDDY-CURRENT CONDUCTIVITY, ONE PART PER LOT, RESULTS REPORTED AGAINST AMS2658 LIMITS.
• PMI READINGS TAKEN ON BARE METAL BEFORE PLATING OR COATING, OR ON A MASKED SPOT.

Decide the sample size from the risk. A mix-up usually happens at the stock level, one wrong bar in a bundle or one wrong bin, so testing one piece per heat catches a wrong heat but not a wrong bar. Refining and petrochemical plants, which typically base their material verification programs on API Recommended Practice 578, often test every alloy component in a pressure boundary. For an OEM machined part, a reasonable middle ground is every bar or casting lot at receipt plus a sample of finished parts, and every part where a wrong grade would be a safety or pressure failure.

Finally, require the report to show the measured values, not just “pass”. A useful PMI record carries the part or lot number, the heat number, the instrument and its check against a reference sample that day, the element values and the operator. That record is what lets PMI connect to the certificate of conformance and the mill certificate rather than sit beside them.

What mistakes make a PMI result worthless?

Reading through plating or coating, trusting the grade label instead of the element values, and accepting XRF for a carbon-defined grade are the three that come up most. Each produces a confident “pass” that answers the wrong question.

Other recurring problems are smaller but real. An analyzer that has not been checked against a reference sample of known chemistry that day can drift. A reading taken on a curved, small or rough surface that does not cover the window gives a low, noisy result. A test on a sample part that was pulled from a different lot than the shipment proves nothing about the shipment. And a PMI report with no heat or lot number cannot be tied to anything. When a reading lands near a specification limit, treat it as a reason for a lab analysis, not as a pass or a reject; handheld precision is coarser than the lab methods the specifications assume. If you are also checking a substitution of lead-bearing brass, the grade logic for water parts is in lead-free brass for NSF 61 and NSF 372.