Delayed weld cracking

Why can a weld that passes every inspection on the day it is made split open two nights later with nobody touching it?

A welder finishes a joint on Friday. It is dye-penetrant tested, radiographed, signed off. Nothing loads it over the weekend, the shop is empty and cold. On Monday there is a crack along the toe of the weld.

That rules out the usual suspects at once. Fatigue needs cycles and there were none; overload needs a load and there was none. Whatever did this was present on Friday and took until Sunday to finish. So the interesting question is not what broke the steel but what took two days.

A delay means something had to move. What could be moving inside a solid steel joint at room temperature over a weekend? Not the iron atoms — this collection's walk-through of work hardening makes the point that at ambient temperature steel has nothing like the mobility needed to rearrange itself. But hydrogen is the smallest atom there is, and it does not sit on lattice sites; it slips through the gaps between iron atoms, diffusing useful distances in hours while everything else in the lattice is effectively frozen. If a delayed process is happening, hydrogen is the only plausible traveller.

Where would it come from? The arc is far hotter than any dissociation temperature, so any hydrogen-bearing molecule near it — water above all, plus oil, paint, and rust, which holds water — is torn into atoms, and molten steel dissolves those atoms readily. The pool is charged with hydrogen from moisture in the electrode coating or flux, from damp plate, from humid air.

Now follow the cooling, because two things change at once and both work against you. First, solubility. Steel at welding temperature is austenite, which holds a lot of hydrogen; as it cools it transforms to ferrite or martensite, which hold very little. The hydrogen has nowhere to go quickly, so the cooled joint is left supersaturated.

Second, microstructure. A weld is a small heated volume attached to a large cold one, so it cools fast, and fast cooling in a steel with enough carbon and alloy produces martensite: hard, strong, low in toughness. That susceptibility is what the carbon equivalent estimates — the IIW formula sums carbon, manganese over six, chromium, molybdenum and vanadium over five, and nickel and copper over fifteen.

Third leg, and it needs no chemistry at all: the joint is in tension. The weld metal solidified hot and then contracted while the surrounding plate held it, and restrained contraction leaves residual tensile stress that can approach the yield strength with nothing applied from outside. That is the "no load" part of the puzzle answered — the load was built in on Friday.

So put the three together and ask what the hydrogen does over the weekend. It drifts toward regions of high hydrostatic tension, because tension dilates the lattice and lowers the energy of a dissolved atom there — and the most triaxially stressed places are the sharp features: the weld toe, the root, an inclusion, the coarse-grained heat-affected zone. Hydrogen accumulates there, locally embrittling the metal until a small crack initiates; the crack tip is a stronger stress concentration still, hydrogen migrates to it, and the crack advances a step, then waits for the next delivery. That stop-start creep is why the failure is slow, and why it completes long after everyone has gone home.

There is a fourth leg, easy to miss: temperature. Too cold and hydrogen barely diffuses; too hot and it escapes faster than it can concentrate. Cold cracking lives in a window around ambient — precisely the temperature the joint reaches a few hours after welding.

Which gives a clean test. If hydrogen is the agent, then removing it and nothing else should stop the cracking. It does: baking electrodes dry, switching to a low-hydrogen process, cleaning the plate, or holding the joint warm for a few hours so the hydrogen leaves before the metal cools into the brittle window all suppress it, with steel, geometry and restraint unchanged. The refuting observation would be equally clear: identical cracks at the same rate on rigorously dried consumables and clean dry plate, or cracks appearing immediately on cooling rather than hours later, or cracking a post-weld hydrogen soak does not touch. Any of those would point elsewhere — to solidification cracking, say, which happens hot and instantly as the last liquid film between grains tears.

The tempting summary is "hydrogen makes steel brittle". Closer is: hydrogen makes a particular steel brittle in a particular state, and only where tension has drawn it. Tough, low-carbon-equivalent weld metal tolerates hydrogen that would crack a hard martensitic zone beside it — which is why the crack so often sits not in the weld but in the heat-affected zone a millimetre away, the hardest, most restrained, least tough metal in the assembly.

It is also worth separating physics from rules. Preheat temperatures, interpass minima and hydrogen-content limits are not physical constants; they are prescriptions derived from carbon equivalent, thickness, restraint, heat input and consumable class, differing between AWS, EN and other standards and by steel grade, so no single number is quotable without saying whose rule it is. The same goes for holding off inspection: some codes require testing of certain high-strength steels to be delayed by a set interval, commonly cited as forty-eight hours, because Friday's clean radiograph is not evidence about Monday. That interval is a code requirement written around the diffusion time, not a property of steel.

The precise mechanism by which dissolved hydrogen lowers local fracture resistance is still argued — whether it chiefly weakens the atomic bonds ahead of the crack, or eases the very dislocation motion that leads to failure. The engineering control does not wait on that argument, because every candidate mechanism needs the hydrogen to be there.

• How hydrogen charging from cathodic protection or pickling embrittles high-strength fasteners that were never welded.

Every term the collection defines is gathered in the glossary.