Hydrogen embrittlement — delayed fracture after plating
Why class 10.9 and 12.9 fasteners crack hours or days after zinc plating, how the preloading test to ISO 15330-1 confirms it, and the ISO 4042 baking rules — ≥ 390 HV, start within 4 hours — that prevent it.
| Defect family | Zinc plating |
|---|---|
| Related service | Zinc Plating Cr3+ |
| Reviewed |
Standards commonly cited
Grades this affects
What you see
Heads pop off or bolts split longitudinally hours to days after plating — often 2 to 48 hours after tightening, sometimes weeks later on the shelf. The fracture is brittle with no necking, usually starting under the head fillet, at the first engaged thread or at a thread root. The same lot passed hardness and tensile before plating, there is no overload evidence on the joint, and the failures cluster in the hardest parts of the lot.
How to confirm it
The recognised acceptance test is the preloading (sustained-load) test to ISO 15330-1: assemble the fasteners on a wedge or fixture at a defined proportion of the minimum tensile load and hold for at least 48 hours — no fracture is the pass. On the parts that already failed, read the fracture face: hydrogen-assisted cracking is intergranular near the origin and turns ductile at final rupture, which separates it from a quench crack (present before plating, oxide-free but not delayed) and from fatigue (beach marks). Then cross-check hardness to ISO 6507-1 / ISO 6508-1 against the class in ISO 898-1, and read the bake record — temperature, time at temperature and, critically, the elapsed time between the end of plating and the start of the bake.
Related calculator: Hydrogen de-embrittlement bake plannerRoot causes — most common first
- 1No de-embrittlement bake at all on parts at or above property class 10.9 / about 390 HV, where ISO 4042 calls for relief; parts in the 320–390 HV band should be assessed rather than assumed safe.
- 2Bake started too late — relief is expected to begin as soon as possible and normally within 4 hours of plating; hydrogen that has already reached traps and opened microvoids is far harder to remove.
- 3Bake too short or too cool for the strength and section — a nominal 4 h at 190–220 °C does not cover heavy sections or parts above roughly 1400 MPa, which need longer holds.
- 4Too much hydrogen charged before plating — aggressive hydrochloric or sulphuric pickling, long acid soaks, or cathodic electro-cleaning on an already hardened part.
- 5Hardness above the drawing — a lot tempered low or run to the top of the class band is dramatically more susceptible; susceptibility rises sharply above about 1000–1200 MPa tensile.
- 6Stress raisers and locked-in stress — unradiused head fillets, threads rolled before heat treatment, deep tool marks and high residual tension concentrate hydrogen exactly where it does the most damage.
Process controls that prevent it
- Decide the bake from strength, not from habit: class 10.9 and above, about 390 HV or roughly 1000–1200 MPa tensile is the trigger band in ISO 4042 unless the drawing says otherwise.
- Start the bake within 4 hours of plating and log the plating-finish time and the oven-entry time on the lot record — the clock is the control, not the recipe.
- Hold at a typical 190–230 °C for 4–24 hours, longer for higher strength and heavier sections, and count time at temperature from when the load reaches temperature rather than from oven switch-on.
- Cut the hydrogen input: mechanical descaling or short inhibited pickling instead of long acid soaks, and no cathodic electro-cleaning on hardened parts.
- Design the risk out at the top end — for class 12.9, springs, case-hardened self-tapping screws and anything above about 1400 MPa, a zinc flake system to ISO 10683 is applied and cured with no electrolytic step, so no hydrogen is introduced.
- Qualify then police the process: run a preloading test to ISO 15330-1 when the process, hardness band or supplier changes, and keep hardness inside the ISO 898-1 band so the lot never drifts to the brittle end.
What to send us for diagnosis
- The fractured parts exactly as they broke — do not clean, tumble, re-plate or wire-brush the fracture face, and wrap each piece separately.
- The drawing or specification: property class, hardness range, coating call-out, any customer standard, plus the material grade and heat-treatment route.
- The process history — pickling type and time, plating time, when plating finished, when the bake started, oven temperature and hold time. The gap between plating and bake is usually the answer.
- A few unbroken plated parts and a few unplated parts from the same lot, so hardness and coating thickness can be compared against the failures.
We reply to quote requests within 24 hours.
Part types this shows up on
Questions engineers ask
Does baking remove all the hydrogen?+
No. A de-embrittlement bake diffuses out mobile hydrogen. It cannot reverse cracking that has already started and it does not empty irreversible traps. That is why the delay before baking is treated as a process control, and why very high-strength parts are usually moved to a non-electrolytic coating instead of being baked harder.
Class 8.8 — do we still need to bake?+
Normally no. Risk climbs steeply above roughly 1000 MPa tensile and 320–390 HV, so standard class 8.8 fasteners are generally outside the requirement. Two exceptions catch people out: case-hardened self-tapping and self-drilling screws, whose surface hardness is far above the core class, and any 8.8 lot that has drifted hard. Check the hardness, not the label.
We baked and it still cracked. What now?+
Look at the four hours first — a bake that starts a shift late behaves almost like no bake at all. Then check the hold, using time at temperature for the actual load rather than the oven set-point, and re-check hardness. If the strength is above about 1400 MPa, or the part is a spring or a 12.9 socket screw, the honest answer is often to stop electroplating it and specify a zinc flake system to ISO 10683.
Guidance based on published standards and general fastener metallurgy. Values are typical ranges that depend on the part, the material and the governing specification — the customer drawing and the applicable standard always prevail. This page does not replace a qualified failure analysis of your own parts.