Hydrogen embrittlement is the rare mechanism whose central physical evidence leaves the part on its own. Atomic hydrogen that has not been trapped irreversibly diffuses out of steel at ambient temperature, so a direct measurement of hydrogen content grows less meaningful with every week the component sits in a bag. What remains is durable but indirect: fracture shape, strength level, microstructure, and what exemplar material from the same lot will do under test.
The evidence that evaporates
Diffusible hydrogen is measured by heating or melting a sample and collecting what comes off, typically by inert gas fusion. Applied promptly to a fractured component it can establish that hydrogen was present at a level consistent with the mechanism. Applied months later it establishes very little, because the hydrogen that mattered has gone.
This is why the handling advice for a suspected hydrogen failure is unusually urgent, and why it is so often overtaken by events. The part goes into a drawer while commercial questions are worked out, and by the time anyone instructs an examination the measurement that speaks directly to the mechanism has decayed.
What the fracture surface retains
Fracture morphology does not decay. Under scanning electron microscopy, hydrogen-assisted cracking in high-strength steel characteristically shows an intergranular path, separation along prior austenite grain boundaries producing the faceted appearance described as rock candy; in some materials it presents instead as quasi-cleavage. Either way, the surface shows little of the microvoid coalescence that marks a ductile overload.
The macroscopic picture supports it: a flat fracture normal to the tensile direction, an origin at a stress concentration, essentially no necking or deformation, and often a small final fast-fracture region. These features survive if the surface is protected from corrosion and handling.
Morphology identifies a family, not a source
Intergranular fracture narrows the field without closing it. Temper embrittlement, grain-boundary segregation, liquid metal embrittlement and stress-corrosion cracking can all produce intergranular separation; separating hydrogen embrittlement from stress-corrosion cracking is a substantive analysis in its own right, treated separately.
Nor does morphology indicate where the hydrogen came from, or when: a fracture face cannot distinguish hydrogen absorbed on a finishing line from hydrogen absorbed in service. An opinion resting entirely on fractography is doing less than it appears to.
Strength level and microstructure
Susceptibility is not uniform across steels. It rises sharply with strength and hardness, which is why the mechanism concentrates in quenched-and-tempered components, spring materials, prestressing wire and hardened parts, and is rare in ordinary structural grades. Untempered or lightly tempered martensite is more vulnerable than a properly tempered structure.
A hardness survey is therefore among the highest-value tests available: quick, inexpensive and comparable against published limits, whether the hardness ceilings in NACE MR0175 and ISO 15156 for H2S service or the limits in the component specification. A part measurably harder than its specification allows is a finding that stands on its own.
What exemplar testing can establish
Testing the failed part tells you about a part that has already fractured. Testing exemplar material from the same lot, with the same processing history, can establish susceptibility, the more useful proposition and the reason unfailed hardware from the batch is worth preserving.
ASTM F519 uses notched specimens under sustained load to qualify plating processes and maintenance chemicals. ASTM F1624 determines a threshold stress for hydrogen-assisted cracking by incremental step loading, producing a measured value rather than a pass or fail. ASTM G142 addresses susceptibility in hydrogen-containing environments at elevated pressure or temperature. Each answers a different question, and choosing wrongly yields a result that is technically valid and forensically irrelevant.
Metallography and the coating cross-section
Sectioning through the fracture origin shows the crack path relative to the microstructure and whether secondary cracks are present. Parallel cracks that never became the primary failure support an environmental or hydrogen mechanism over a single overload event.
Where the component is plated, the same cross-section documents coating thickness and continuity and whether corrosion product sits within the crack. Product inside a crack points toward an environment acting during propagation; a clean crack beneath an intact coating points toward hydrogen already present when load was applied.
What no test can recover
Several questions have no retrospective answer and should not be presented as though they do. How much hydrogen was in the part at assembly cannot be measured later. The moment the crack initiated cannot be read off the fracture surface. Whether a particular bake would have prevented this particular failure is a counterfactual exemplar testing can inform but not resolve. Stating those limits is not a weakness; a confident figure with nothing behind it is the finding most likely to be dismantled.
Sequencing the examination
Order matters, because several of these techniques are destructive and one is time-sensitive. Photography and non-destructive examination first, then hydrogen analysis if timing permits, then fractography on the undisturbed surface, then hardness and chemistry, then sectioning and metallography last. Cleaning a fracture surface aggressively to see it better is a common and irreversible error.
Mating halves, adjacent hardware and unfractured items from the same lot deserve the same treatment. They are frequently more informative than the broken piece.
Where these opinions are challenged
Predictably: that intergranular fracture was read as proof of hydrogen without excluding the alternatives; that a hydrogen measurement taken long after the failure was given weight it cannot carry; that exemplar testing used material with a different processing history; that hardness was measured somewhere unrepresentative of the origin.
An analysis that records when each specimen was taken and in what condition, states what each test can and cannot support, and identifies the competing mechanisms set aside and why, is materially harder to displace.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.