1. A failure that is regularly misdiagnosed as "bad material"
Start with a specific scene.
A batch of class 12.9 bolts. Goods-in inspection passes. Assembly torque is normal, no thread stripping, no overload. Then, within 24 to 72 hours of assembly, several of them snap — at the head or at the thread — under a load far below the rated capacity, in some cases while merely holding preload. Take one apart and look at the fracture: flat, clean, almost no necking.
The usual response is to declare the material non-conforming, change supplier, and see the same problem return in the next batch. Because the real cause is not in the heat number from the steel mill — it is in the surface finishing process.
To identify hydrogen embrittlement, remember that three features occurring together are highly suspicious: fracture at a stress far below the strength limit, failure that is late (hours to days after assembly), and a brittle, flat fracture surface.
2. The three necessary conditions for hydrogen embrittlement
Hydrogen embrittlement is not caused by a single factor. It is the result of three conditions being met simultaneously:
| Condition | What it means | How to judge it in practice |
|---|---|---|
| A susceptible material | The higher the strength or hardness of the steel, the more sensitive it is to hydrogen | Tensile strength Rm ≥ 1000 MPa, corresponding to roughly ≥ 320 HV — in practice class 10.9 and above |
| Hydrogen ingress | Hydrogen atoms enter the steel and diffuse within it, either during manufacturing or in service | Presence of wet processes such as acid pickling or electroplating (see the list in section 4) |
| Tensile stress | A sustained tensile stress: assembly preload plus working load plus residual stress | A tightened bolt already carries a high tensile stress, so this condition is almost automatically satisfied |
The point of the three conditions is this: remove any one of them and hydrogen embrittlement cannot occur. So there are three lines of countermeasure — make the material less susceptible, cut off the hydrogen source, or reduce the tensile stress. In engineering practice the second is the most effective and the most controllable.
One concept needs separating out. What is discussed here is internal hydrogen embrittlement (IHE): hydrogen introduced during manufacturing that causes delayed fracture after assembly. Brittle fracture caused by hydrogen re-entering the part in service — from cathodic protection or a corrosion reaction — is normally classified as stress corrosion cracking (SCC). The mechanism is similar, but the responsible stage in the chain is different.
3. Two hardness thresholds: 320 HV and 390 HV are not the same thing
This is where reference material most often gets muddled. The ISO 4042 framework actually draws three lines, and they mean completely different things:
| Threshold | Value | What the standard says | What you have to do |
|---|---|---|---|
| Risk threshold | Rm ≥ 1000 MPa (≈ ≥ 320 HV) | Fasteners manufactured to that strength class, carrying tensile stress and electroplated, are at risk of failure by hydrogen embrittlement | Assess the risk and take measures to reduce it; this is where it starts to matter |
| Mandatory baking threshold | Hardness / core hardness > 390 HV (≈ ≥ 40 HRC) | Above this hardness, de-embrittlement baking is mandatory for electroplated fasteners; below it, the decision is risk-based | Baking is not optional; it must be performed and recorded |
| Written-agreement threshold | Hardness > 365 HV | Class 10.9 and 12.9 fasteners in that band (i.e. those at the upper end of the permitted hardness range) should be covered by a written agreement between purchaser and supplier on how hydrogen risk is managed | Ask the supplier for a hydrogen-control plan and test records, not just a system certificate |
There is also a sentence from ISO 4042:2018 worth reading closely, because it states the nature of the problem precisely: class 10.9 fasteners correctly manufactured to the intended material and metallurgical properties in accordance with ISO 898-1 will not fail by internal hydrogen embrittlement and do not require baking.
That sentence is not a reason to relax — it tells you where the problem lies: the root of hydrogen embrittlement is loss of process control, not the property class itself. If the raw material and heat treatment are both sound, class 10.9 does not inherently "need baking". The reason so many class 10.9 parts are baked in real production is that baking is being used to cover loss of control in pickling and plating. Baking is a remedial measure, not a guarantee.
4. Where the hydrogen comes from: a process-by-process source list
| Process | Introduces hydrogen? | Notes |
|---|---|---|
| Acid pickling (descaling, derusting) | Yes | The acid reacts with the steel and liberates hydrogen, some of which diffuses into it. What is often overlooked is that pickling sits upstream of many other processes |
| Electroplated zinc / nickel / cadmium | Yes (the dominant source) | The cathodic process itself evolves hydrogen. In real production, most hydrogen contamination of fasteners comes from electroplating |
| Cathodic electrocoating, electrolytic cleaning | Yes | Also cathodic hydrogen evolution, and easily missed |
| Phosphating | Yes | Phosphating is normally preceded by acid pickling; the pickling is the hydrogen source |
| Hot-dip galvanising | Medium / low | Hydrogen comes mainly from the upstream pickling and has to be assessed and controlled; the hot-dip bath itself is hot (about 440–470°C) |
| Mechanical zinc plating | Essentially none | A mechanical/chemical process: degreased parts are tumbled with zinc powder and glass beads, and the beads peen the zinc onto the surface. No electrolysis involved |
| Non-electrolytic zinc-flake coating (zinc-aluminium flake, sold under trade names such as Dacromet) | Essentially none | Applied mechanically, virtually eliminating the possibility of hydrogen embrittlement; resistant to 300°C; suitable for M4 and above |
| Black oxide / stainless steel blackening | Low | A chemical conversion process with no cathodic hydrogen evolution; but it reduces the corrosion resistance of A1–A4 stainless steels |
The conclusion from this table is straightforward: any wet process that uses electrical current is a hydrogen source; any mechanical application process that does not, is not. That is exactly why the alternative process routes cluster around mechanical zinc plating and zinc-flake coatings.
5. Delayed fracture: the time window and the fracture surface
The nastiest thing about hydrogen embrittlement is that it is delayed. It does not happen on the assembly line, so assembly records, torque traces and operator observation all fail to catch it.
| Feature | Appearance | Diagnostic value |
|---|---|---|
| Time to fracture | Mostly hours to days after assembly | Clustered in time and lagging the assembly moment — the strongest single clue |
| Stress level | Stress at fracture is far below the tensile strength, sometimes only the preload | Rules out overload and rules out torque runaway |
| Fracture morphology | Brittle, no necking, flat and clean face | Stands in sharp contrast to a ductile overload fracture, which shows obvious plastic deformation and a rough face |
| Initiation site | Usually at a stress concentration: first engaged thread root, head-to-shank fillet, a scratch or surface defect on the shank | Points to stress concentration rather than insufficient bulk strength |
| Section effect | Brittleness increases as the cross-section decreases | Fine-thread and small-diameter high-strength bolts are more dangerous |
A very practical diagnostic rule: if several failures occur in the same batch, from the same process route, in the same assembly state, and they cluster within a few days, suspect hydrogen embrittlement first rather than "a bad batch of steel". A genuine material problem usually produces a more random failure distribution with a weaker link to batch identity.
6. De-embrittlement baking: temperature, duration, timing
Where electroplating cannot be avoided, baking (de-embrittlement treatment, also called hydrogen relief) is the most common and most effective remedial measure. The parameters are as follows:
| Parameter | Value | Notes and constraints |
|---|---|---|
| Temperature | 190–230°C, commonly 200–230°C in practice | Must be below the original (first) tempering temperature of the part, otherwise the mechanical properties already achieved are destroyed. Electroplated zinc parts are normally baked at no more than 220°C; electroplated cadmium parts normally no more than 200°C; certain coatings (tin, for example) and certain part characteristics also impose an upper limit |
| Duration | 4–24 h | For susceptible fasteners (hardness / core hardness above 390 HV), 8–10 h at 190–220°C is the minimum recommended baking duration for electroplated zinc parts. The higher the strength and the thicker the section, the longer the time needed, up to 24 h |
| Start timing | Begin within 4 h of plating, the sooner the better | If the parts only reach the oven after 4 h, irreparable microcracks may already have formed — local residual stress drives hydrogen toward the high-stress regions |
| Position in the sequence | Plate → bake → conversion coating / sealant / topcoat | Baking usually takes place after plating and before conversion coatings and/or sealers and/or topcoats. Other sequences may be appropriate for other coating systems |
| Upper constraints | Must not exceed the original tempering temperature; must not damage coating performance | Exceeding the temperature or time also affects the beneficial compressive residual stress produced by thread rolling after heat treatment |
It is worth understanding what baking actually does. It is not simply "driving the hydrogen out". It does two things:
- Diffuse as much free hydrogen out of the part as possible;
- Allow the remaining hydrogen to migrate to, and be trapped at, "traps" in the steel — inclusions, grain boundaries and similar sites — so that it can no longer migrate freely into high-stress regions.
Research indicates that what governs the outcome is the local hydrogen concentration, not the total hydrogen content of the part. That also explains why baking significantly reduces risk but cannot fully eliminate it. ISO 4042's own wording is that complete elimination of hydrogen embrittlement cannot be guaranteed.
The oven itself also needs controlling: loading arrangement (parts must not be piled up; the hot air has to reach every piece), soak time in the oven, and temperature uniformity.
7. Why "only 4 hours" can actually make things worse
This section is the most counter-intuitive and the most useful part of the article.
ISO 4042:2018 states explicitly that the common practice of baking electroplated zinc fasteners at about 190°C is inadequate to extract hydrogen, because the zinc layer itself is an effective barrier to hydrogen diffusion. The result is that a baking duration of 4 hours may even be harmful and can lead to sporadic failures. For baking to be effective and beneficial, the duration should be extended.
So where does the figure of "4 hours" come from? Its actual role is closer to a workable time frame and a quality-assurance mechanism (something that makes it possible to monitor whether the procedure was followed) than to a physically sufficient condition. The standard also adds a specific piece of discipline:
- The time interval between plating and baking should not be used as a rigid criterion for the acceptability of a fastener batch; still less should it be used as a basis for allocating the root cause of a failure.
That sentence is worth printing out and pinning up in the quality department. It means you cannot relax the remaining checks because the supplier says "we had them in the oven 2 hours after plating", and equally you cannot pin the root cause on the fastener simply because the interval on the failed parts exceeded 4 hours.
Note also how this relates to other standards: the baking requirements in ISO 2081, ISO 9588 and ISO 19598 are too broad and are not applicable to fasteners. Judging de-embrittlement of fasteners should be done against a fastener-specific standard such as ISO 4042.
8. The more reliable route: avoid electroplating at the source
Since baking is only remediation, the most reliable strategy is simply not to introduce hydrogen in the first place. For different target classes, the order of preference among alternative routes is:
| Target class / scenario | Recommended process | Why |
|---|---|---|
| Class 10.9, 12.9 (especially the high-hardness end) | Non-electrolytic zinc-flake coating (zinc-aluminium flake) > mechanical zinc plating > inorganic zinc coating | The application process generates no hydrogen, virtually eliminating the possibility of hydrogen embrittlement; zinc-flake coatings also withstand 300°C and suit M4 and above |
| Class 12.9 | Mechanical zinc plating or zinc-flake coating first; avoid electroplated zinc, and avoid hot-dip galvanising too | Class 12.9 sits in the highest-risk band for hydrogen embrittlement; hot-dip galvanising is not electrolytic, but the upstream pickling still introduces hydrogen |
| Where electroplating is genuinely unavoidable | Tighten the baking (upper end of the temperature range, longer duration) + hydrogen testing per batch or at a defined frequency + a written purchaser-supplier agreement | Manage the risk as "known and controlled" rather than assuming "it should be fine" |
| Where the material can be changed | Switch to stainless steel or an acid-resistant grade | One way to cut the hydrogen source. But note: high-strength stainless steels (martensitic, precipitation-hardening) are themselves at risk of hydrogen embrittlement and stress corrosion cracking — not an unconditional safe exit |
9. How to verify that de-embrittlement worked
Once baking is done, how do you know it achieved anything? ISO 4042 specifies that where the core or surface hardness of the fastener is above 320 HV (corresponding to classes 10.9 and 12.9), hydrogen embrittlement should be detected by test.
The usual method is a preload sustaining test: tighten the fastener to near or at yield, hold it in that state for 48 hours, and pass it if no fracture occurs and no visible cracks are found. ISO 15330 is the standard written specifically for this preload test for the detection of hydrogen embrittlement in fasteners.
There are two further approaches: process-control verification for the plating line itself (long-duration sustaining tests on notched, high-hardness specimens to verify the hydrogen control level of the whole line), and batch sampling.
For purchasing requirements, write it like this: require the supplier to provide hydrogen embrittlement test records per batch or at a defined frequency, not a quality management system certificate. A system certificate proves that a procedure exists; a test record proves that this particular batch is not at risk.
10. Three supplementary lines of defence on the design side
Even with the process side fully under control, there are three things design can do to push the risk down further:
- Reduce stress concentration. Hydrogen always migrates to high-stress regions. Avoid sharp corners, avoid abrupt thread run-out, avoid scratches and machining marks on the shank, and make transition fillets as generous as possible. Failures start at these locations not because they are "weak" but because the local stress there attracts hydrogen.
- Control the preload. Tensile stress is one of the three necessary conditions, and preload is its main source. On joints that are sensitive to hydrogen embrittlement, do not push assembly stress to the limit — which conflicts with the usual "maximise preload to prevent loosening" instinct, so it has to be weighed against how critical the joint is.
- Avoid re-introducing hydrogen in service. Watch the combination of cathodic protection, acidic media and corrosive environments; failures in that setting are closer to stress corrosion cracking.
11. Five common mistakes
| Mistake | Consequence |
|---|---|
| Treating "higher property class is better" as the default | Class 12.9 plus electroplating is one of the most dangerous combinations there is. Class 8.8 is sufficient for most joints, with friendlier toughness and risk |
| Specifying only "zinc plated" without specifying the process | Electroplated zinc, mechanical zinc and non-electrolytic zinc-flake coatings are three completely different things as far as hydrogen embrittlement is concerned, with large differences in cost and risk. Leaving it unspecified hands the choice to the lowest bidder |
| Assuming that baking makes it foolproof | Baking significantly reduces risk but does not guarantee elimination, and baking that is poorly executed can itself be harmful |
| Treating "in the oven 2 hours after plating" as evidence of conformity | That interval is a process-monitoring indicator; the standard states clearly that it is not an acceptance criterion and cannot be used to determine the root cause of a failure |
| Judging delayed fracture as non-conforming material, changing the steel but not the process | The problem is guaranteed to recur. The root of hydrogen embrittlement is in the process route, not in the heat number |
One sentence to close on: managing hydrogen embrittlement is not a matter of one process step but of a whole chain — material and heat treatment set the susceptibility, surface finishing sets how much hydrogen enters, baking sets how much can be recovered, and testing tells you whether you actually know your own state. Lose any one of the four links and the chain is unreliable.
Sources
- ISO 4042, Fasteners — Electroplated coating systems: the hydrogen embrittlement risk threshold (Rm ≥ 1000 MPa / ≥ 320 HV), the mandatory baking threshold (> 390 HV), the written-agreement threshold (> 365 HV), baking temperature/duration/timing requirements, coating thickness limits, and the statement that complete elimination of hydrogen embrittlement cannot be guaranteed.
- ISO 898-1: property classes and hardness ranges (the hardness bands corresponding to classes 10.9 and 12.9).
- ISO 15330, Fasteners — Preload test for the detection of hydrogen embrittlement: tightened to near or at yield, held 48 h, pass criterion of no fracture and no visible cracks.
- ISO 2081, ISO 9588, ISO 19598: noted as having baking requirements too broad to be applicable to fasteners.
- Corporate technical manual: comparison of surface finishing processes (hot-dip galvanising, mechanical zinc plating, non-electrolytic zinc-flake coating, black oxide, phosphating, drying) and their implications for hydrogen embrittlement risk.
- Industry technical literature: the mechanism of hydrogen embrittlement and delayed fracture in electroplated parts, the trapping and redistribution of hydrogen during baking, and the finding that local rather than total hydrogen concentration is the controlling variable.
👉 Engineering services ｜ 👉 Request a technical quote