The crack that shows up tomorrow morning. Hydrogen cracking is the delayed failure of the weld heat-affected zone — it needs hydrogen, a hard microstructure, and stress, and it can appear hours after the weld has gone cold. Every prevention measure attacks one of those three.
Hydrogen cracking is a cold cracking mechanism: a crack that forms in the heat-affected zone (HAZ) of hardenable steels — and occasionally in the weld metal itself — after the weld has cooled, sometimes minutes and often hours later. It is called "delayed" for a reason: the weld can look perfect at shift end and be cracked by morning.
It is the most dangerous weld defect because it is invisible when it forms and it is usually found by failure, not by chance — and because it is entirely preventable with dry consumables, preheat, and controlled cooling. Codes take it seriously enough to schedule delayed inspection: MT or UT at 24–48 hours after welding on critical hardenable-steel joints.
Cold cracking, delayed cracking, hydrogen-induced cracking (HIC), hydrogen-assisted cracking (HAC), underbead cracking, toe cracking, root cracking, fish eyes.
The telling detail is the timeline: no crack at shift end, an indication 24 hours later. If the crack "wasn't there yesterday," it is hydrogen.
Three conditions must ALL be present: hydrogen in the weld, a hard crack-sensitive microstructure in the HAZ, and tensile stress. Remove any one, and hydrogen cracking cannot occur.
Step 1 — The hydrogen arrives. Moisture in the electrode coating, the flux, the plate surface, or the air is broken apart in the arc into atomic hydrogen, which dissolves readily in molten steel. Some escapes as gas — the rest stays dissolved in the weld metal. As the weld cools, steel's ability to hold hydrogen collapses, and the hydrogen is rejected from the solidifying weld metal into the still-hot HAZ. The HAZ is flooded with hydrogen.
Step 2 — The HAZ goes hard. If the HAZ cools faster than the critical cooling rate, austenite transforms to martensite — hard, strong, and essentially brittle to hydrogen. Hardness above roughly 350 HV is the danger zone. Fast cooling comes from thick sections that sink heat, low preheat, low heat input, and cold ambient temperature. This is the step preheat controls: preheat slows the cooling rate and lets the HAZ transform to a tougher microstructure instead.
Step 3 — Stress and the delay. The weld sits under tensile stress — shrinkage restraint, joint fit-up, service load. Hydrogen diffuses through the lattice and concentrates at stress raisers in the hard HAZ: toe notches, slag lines, inclusions. Cracking begins when the local combination of hydrogen, hardness, and stress passes the threshold — and it takes time for hydrogen to diffuse and concentrate, which is why the crack is delayed by minutes, hours, or days at room temperature. Once started, the crack grows stepwise along the HAZ until the stress is relieved or the hydrogen is exhausted.
From moisture in flux, electrodes, plate, and air; also oils and organic contamination. Dissolved in the weld, rejected into the HAZ as the weld cools.
Martensite or bainite in the HAZ, formed when cooling outruns the critical rate. Thick sections, low preheat, and cold weather all push the HAZ hard.
Restraint and residual shrinkage stress — highest at toes, roots, and joint ends, exactly where hydrogen collects. All three together, at the same moment, equals a crack.
The susceptible jobs are the high-profile ones: high-strength low-alloy steels (A514, HY-80), thick sections, high-restraint joints like box closures, and cold-weather work. Not every hydrogen weld cracks — some joints bake the hydrogen out harmlessly — but on these jobs, the margin between harmless and cracked is the procedure.
Control all three conditions and the crack cannot happen. The WPS is written around exactly this — follow it.
Dry the consumables: low-hydrogen electrodes (H4/H8 class) stored in a rod oven at 250–300°F, re-dried per spec (7018 typically 650–700°F for 1 hour) and used within the exposure time. Clean and dry the joint — no moisture, oil, or paint. Check the forecast: cold and humid means preheat is not optional.
Preheat to the WPS temperature — measured at the joint, not the plate edge — and hold it. Use the heat input range the WPS specifies; low heat input invites a hard HAZ. Keep the arc tight and consumables out of their boxes for the minimum time.
Hold the interpass temperature — don't let a hardenable joint cool to room temperature between passes. Clean slag and the root before the next pass. Where the WPS calls for postheat (a hydrogen bake-out at 200–300°C) or a controlled cooldown under insulation, do it and record it.
The "it's a small weld" excuse. Tack welds and partial joints cool the fastest and crack the most — preheat applies to them too. And never weld over a suspected delayed crack: it will re-open through the repair.
A delayed crack is a full excavation job — the visible crack is only the tip. Treat every step as if the crack is longer than it looks, because it is.
| Step | Action | Why |
|---|---|---|
| 1 | Do not weld over it. Mark the crack and define the excavation zone. | Hydrogen cracks run ahead of the indication; welding over one just buries it. |
| 2 | Excavate the crack — grind or gouge to sound metal, and keep going past the last crack tip. | Crack tips run ahead of the visible line; a buried tip re-opens the repair. |
| 3 | Verify by MT or PT on the excavated groove — zero indications anywhere. | Invisible hydrogen cracks hide in the groove walls. |
| 4 | Decide severity — localized cracks repair; large, multiple, or through-thickness cracks may scrap the joint. | The code limits repair scope; repeated repairs on hardenable steel are worse than none. |
| 5 | Re-weld to a low-hydrogen WPS — dry electrodes, preheat, interpass control — and schedule 24–48 h delayed inspection. | The repair must not recreate the conditions, and the verification must catch it if it does. |
Never "burn over" a delayed crack. The crack will propagate through the repair at the same stress raiser, deeper. If the crack extends through the thickness or into the base metal beyond the weld, the component may be beyond repair and must be scrapped. This is why delayed inspection exists: find it before it finds you.
Hydrogen enters through the consumable, the joint, or the air — each process has its own doorway.
The classic source. Cellulosic electrodes (6010/6011) deposit hydrogen with the root — fine for mild steel, dangerous under a high-strength fill. Basic low-hydrogen electrodes (7018, 7018M) must come dry from the oven and stay dry; a damp 7018 is worse than a 6010.
The flux core absorbs moisture in humid shops; a wet spool produces underbead cracks. Gas-shielded FCAW controls hydrogen better than self-shielded, but neither survives a damp spool or a wet joint.
Flux moisture is the hydrogen source; recycled flux carrying rust and fines adds more. Re-dry flux per the manufacturer's spec and respect the reuse limits — this is the process where hydrogen cracking is most often a flux-handling failure.
Low-hydrogen by nature — the tungsten and gas introduce almost none. Hydrogen here comes from contamination: moisture, oil, and grease on the wire or the joint. On hardenable steel, treat cleanliness with the same discipline as electrode drying.
Same story as TIG: clean wire, clean joint, and hydrogen is a non-issue. The classic failure is welding over wet or oily plate — the contamination loads the weld with hydrogen and the hardenable HAZ does the rest.
When hydrogen cracking appears, the first questions are about temperature and moisture — not amperage.
| Parameter | Check | Typical Fix |
|---|---|---|
| Preheat temperature | WPS minimum, measured at the joint before and during welding | Heat to spec (150–400°F by grade and thickness); verify with a contact thermometer |
| Interpass temperature | Never below the preheat minimum between passes | Reheat before continuing; monitor with a pyrometer |
| Heat input | kJ/in or kJ/mm within the WPS range | Low heat input → hard HAZ; raise current or voltage, or slow travel, to spec |
| Electrode moisture | Oven at 250–300°F; re-dry per spec; use within exposure time | 7018: 650–700°F × 1 h; discard electrodes past their exposure time |
| Hydrogen class of consumable | H4 / H8 / H16 diffusible hydrogen rating | Use H4 or H8 for hardenable steel and high-restraint joints |
| Cooling rate / cooldown | Controlled cooldown or postheat required by the WPS | Insulate the joint or postheat at 200–300°C to let hydrogen diffuse out |
Use the Heat Input Calculator to verify the WPS range, and the Filler Metal Selector to pick the right hydrogen class (H4/H8/H16) for the job.
The ovens do the preventing. Consumable handling is where hydrogen cracking is won or lost.
Electrode ovens — storage ovens at 250–300°F, re-dry ovens at 650–700°F, and portable rod boxes — keep low-hydrogen electrodes dry on the floor. Preheat equipment (resistance heaters, induction, torches) matters less than measuring: a contact thermometer or pyrometer at the joint is the real control. Machines with accurate heat-input readouts help you hold the WPS numbers through every pass.
Hydrogen cracking is found by method, by timing, or by failure. Codes schedule the method and the timing for a reason.
Acceptance depends on the code and project: AWS D1.1, ISO 5817, API 1104 and project specifications each set their own limits for this discontinuity. Check the governing document before judging a weld acceptable — there is no universal pass/fail rule.
The questions welders actually ask about hydrogen cracking — answered straight.
It can start minutes after the weld cools and continue for 24–48 hours or more at room temperature — which is why codes require delayed inspection (MT or UT at 24–48 hours) on critical hardenable-steel welds. The delay is exactly what makes it dangerous: the weld looks fine when it leaves the shop.
Any steel that can form a hard heat-affected zone: high-strength low-alloy steels like A514 and HY-80, quenched-and-tempered grades, and thicker sections of ordinary carbon steel. The WPS decides — if the code calls for preheat on your material and thickness, it is because the steel can crack without it.
Never. The crack will run ahead of the repair — through the new weld — because the hydrogen, the hard HAZ, and the stress are still there. Excavate the crack completely, verify removal with MT or PT, then re-weld with a low-hydrogen procedure and fresh preheat.
Yes, if the spec takes them seriously. E7018 is a low-hydrogen electrode by design — its coating must stay dry to stay low-hydrogen. Store at 250–300°F in a rod oven, re-dry per the manufacturer's spec after exposure (typically 650–700°F for one hour), and use within the exposure time, usually four hours. A damp 7018 is a hydrogen factory.
That is the delayed signature: hydrogen from a damp electrode or wet joint, a heat-affected zone that cooled too fast into martensite, and shrinkage stress doing its work while the weld sits. Check electrode storage, preheat, and interpass control — the crack happened because at least one of the three conditions was present.