Why Hydrogen Is the Enemy in a Weld

Why this matters

A weld that passes every check on Friday and is cracked on Monday was not badly welded on Monday. Hydrogen cracking is the one weld defect with a delay built into its mechanism: the hydrogen has to diffuse to where the stress is before anything opens, and that takes hours to days at shop temperature. So the inspection you did while the joint was still warm proved that no crack existed yet, which is a different statement from the one everyone read it as. That single property, delay, is why this failure survives good shops with good welders and good inspection.

Before any of this happens: the arc itself burns unprotected eyes and skin at a distance, so use the filter shade for the process and current with screens around the work per 29 CFR 1910.133 and 29 CFR 1910.252(b)(2). The fume is an inhalation hazard whose constituents come from the base metal and its coating rather than from how the smoke looks, so local exhaust at the arc plus respiratory protection under a written program meeting 29 CFR 1910.134 is the control, not a glove. Two contamination sources in this article carry their own routes: never strike an arc where chlorinated degreasing solvent vapour can reach it, because the ultraviolet from the arc decomposes those vapours into phosgene and other irritants, so degrease with a non-chlorinated product and let the part dry and ventilate before welding; and grinding old paint off before welding releases whatever pigment it carried, commonly lead under 29 CFR 1910.1025 or a chromate primer under 29 CFR 1910.1026, which needs tool-mounted capture or respiratory protection under 1910.134 rather than eye protection alone.

Where the hydrogen comes from

Hydrogen does not need a hydrogen source. It needs a hydrogen-bearing compound in the arc, and water is the commonest one on earth. Every item on this list is ordinary shop material:

• Moisture absorbed into an electrode coating or a flux, which is a function of humidity and exposure time. The related card on electrode storage owns that path.
• Humidity in the air around the arc, which is why the same procedure that never cracks in January cracks in July in a coastal shop.
• Oil, grease, cutting fluid and drawing lubricant, including the lubricant left on solid wire from manufacture.
• Paint, primer and any organic coating.
• Rust, which is a hydrated oxide, and mill scale that has been rained on.
• Shielding gas above its specified dew point, or a wet gas line.

Water in the arc dissociates. The pool dissolves hydrogen readily while it is liquid, and its solubility collapses at solidification, so most of it is rejected into the surrounding solid metal rather than escaping upward.

The three-factor model

Hydrogen cracking needs all three of these at once, and removing any one of them prevents it. This is the operative device in the whole subject, and it is the reason the same electrode is harmless on one job and cracks the next.

1. Diffusible hydrogen in the deposit, above a level the material can tolerate.
2. A susceptible microstructure, meaning a hard, low-toughness structure in the weld metal or the heat-affected zone. On hardenable steels that is set by cooling rate, which is set by preheat, interpass temperature, heat input and section thickness. The interpass card owns those controls.
3. Tensile stress, which is present in every weld from solidification shrinkage and is multiplied by restraint: a member fixed at both ends, a thick section, a joint welded between two already-welded assemblies.

Two of the three are always partly present. Every weld has some residual tension, and every carbon steel weld has some hydrogen. So in practice the question is never "is hydrogen present" but "which of the three can I move furthest on this joint."

Why it is delayed, and how long

Diffusion takes time and it is temperature-dependent. Hydrogen migrates through the metal to regions of high triaxial stress and to the hardest structure it can find, and the crack opens when the local concentration and the local stress cross a threshold together. On a cold, thick, restrained joint that can be hours; it can also be a day or two. AWS D1.1, in the edition your project specification or adopted building code invokes, recognises this directly by requiring visual inspection of certain quenched and tempered steels to be delayed at least 48 hours after the weld is completed. Your shop's own hold period is an operational decision, but the reason for it is not.

The practical consequence: any inspection performed on a hardenable steel before the hold period has expired is a check on workmanship, not a check for hydrogen cracking, and it should be recorded that way rather than as a clean joint.

Telling it apart from the crack it gets confused with

Solidification cracking, the hot crack, is the one people mistake it for, and the two want opposite corrections.

Slowing down helps one and can worsen the other. Read the crack before you change the procedure.

Worked example: cracked over a weekend

A shop welds a brace into an existing frame on a Friday afternoon in a humid August. The brace is 1 in plate on a hardenable low-alloy steel, welded on both sides, between two members that are already fixed. Visual and magnetic particle inspection at the end of the shift: clean. Monday morning a crack runs along the weld toe on one side, roughly the full length.

Candidates eliminated on evidence.

• Fit-up or lack of fusion. Would have been present Friday and visible in the same magnetic particle check that passed. It is also a root or interface feature, and this crack is at the toe in parent metal.
• Hot cracking. Centreline and crater are the signatures, and it opens while the metal is freezing, so it would have shown Friday. This one did not exist Friday.
• Overload. Nothing was loaded over the weekend, and the frame is not in service.
• Fatigue. Requires cycles that have not happened.

That leaves the mechanism whose defining property is the delay. Now test all three factors against what was actually done.

Factor 1, hydrogen. The carton of low-hydrogen electrodes was opened Monday of that week and left on the bench in an unconditioned shop for the whole week. That is roughly 108 hours of continuous atmospheric exposure, counted as elapsed time out of the container rather than as shift hours, because a coating on a bench absorbs moisture overnight too, against a limit for that classification measured in a small number of hours. Present.

Factor 2, microstructure. Ambient in the bay at 5 pm was around 70 degrees F, and no preheat was applied because the WPS was not pulled. On 1 in hardenable plate the section is thick enough to draw heat away fast in three dimensions. Present.

Factor 3, restraint. The brace was welded between two fixed members with no free end and no sequencing plan, so all the shrinkage of both fillets went into stress rather than movement. Present.

Three of three. This joint was not unlucky, it was arithmetic.

What the repair changed. The shop removed two of the three factors rather than one, on the reasoning that restraint was the one they could not remove on an existing frame.

• Hydrogen: fresh electrodes of an H4 classification issued from the holding oven by the rod into quivers, with unused rods returned within the exposure limit for that classification. H4 means a maximum of 4 millilitres of diffusible hydrogen per 100 grams of deposited weld metal, measured on the test that classification specifies, on the electrode in the condition the specification defines. Correction, printed: that 4 is a property of the electrode as tested and packaged, not of the rod in the welder's hand after a week on a bench, which is exactly what went wrong the first time.
• Microstructure: preheat and interpass minimum taken from the WPS and confirmed with a contact pyrometer before every pass, six passes, six readings.
• Restraint: unchanged, and recorded as unchanged.

Confirmation. The repair was welded Tuesday. Magnetic particle inspection was performed after a 72 hour hold rather than at the end of shift, and the joint was clean. Three further identical braces were welded the same way over the following two weeks with the same hold and the same result, giving four clean joints against one cracked one.

The honest limit on that result. Four joints is a count, not a rate, and two variables moved together, so this does not establish which of the two corrections did the work. It establishes that the pair is sufficient on this joint. If the shop wants to know whether preheat alone would have carried it, that is a separate test they have not run.

Check it against the rules the siblings state

• Delay honoured with a figure: the failed joint was inspected at end of shift on the day of welding, about 2 hours after the arc; the repair was inspected after 72 hours, against the 48 hour delayed-inspection precedent AWS D1.1 sets for certain quenched and tempered steels in the adopted edition.
• Three-factor test printed per factor, not asserted: hydrogen present at roughly 108 hours of elapsed atmospheric exposure; microstructure susceptible at 1 in section with 70 degrees F ambient and zero preheat; restraint present with both ends fixed. Three of three, and the repair removed two.
• Consumable claim carries its condition: the H4 figure of 4 millilitres per 100 grams is stated as measured on the classification's test in the specified condition, and the example says outright that the bench-aged rod no longer carried it.
• Interpass floor obeyed in the example: six passes, six contact-pyrometer readings taken before striking, per the WPS the repair was run to.
• Comparator stated as uncorrected: four clean joints against one cracked joint is reported as a count with two variables moved, not as a rate and not as attribution to preheat.

References

• AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the adopted building code, including its delayed visual inspection requirement for certain quenched and tempered steels
• AWS A5.1 (carbon steel covered electrodes) and A5.5 (low-alloy covered electrodes) for the optional diffusible hydrogen designators and the test condition behind them, as invoked by your WPS or contract
• 29 CFR 1910.252(b)(2) and 1910.133 for arc radiation; 29 CFR 1910.134 for respiratory protection; 29 CFR 1910.1025 (lead) and 1910.1026 (hexavalent chromium) for coating removal before welding
• See related: What a Welded Joint Changes About the Parent Metal; What Interpass Temperature Is Protecting; What Electrode Storage Has to Do With Cracking