A crack is a planar separation that grows under stress — the only weld discontinuity that propagates on its own. Hot cracks form as the pool freezes; cold cracks can appear days after the weld has cooled.
A crack is a fracture-type discontinuity: a planar separation within the weld metal, the heat-affected zone, or the base metal adjacent to the weld. It forms when tensile stress exceeds the material's capacity to deform — during solidification, during cooling, or days later as hydrogen does its work.
Cracks are the one family of defects that grow under load — and nearly every code rejects them outright. AWS D1.1, ISO 5817, and API 1104 prohibit cracks of any size, which puts them above every other discontinuity in consequence. A crack also signals a metallurgy problem — steel chemistry, hydrogen, cooling rate, or restraint — not just a technique slip.
Hot cracks, cold cracks, hydrogen-induced cracks, crater cracks, centerline cracks, transverse cracks, longitudinal cracks, toe cracks, root cracks, check cracks.
Crack names describe where they live and which direction they run. Name it first — the name tells you the mechanism.
Every crack needs ingredients in combination: a susceptible microstructure, a source of stress, and — for cold cracks — hydrogen. Match the crack to the category below.
Cracks are named two ways. By orientation: a longitudinal crack runs along the weld axis, a transverse crack crosses it. By location: centerline, toe, root, crater. Add the timing — hot or cold — and you have the diagnosis. A hot crack forms while the weld is still hot. A cold crack forms hours to days later, at or near room temperature. They are different diseases with different cures.
Form while the pool freezes. Low-melting films of sulfur, phosphorus, or copper segregate to the grain boundaries; shrinkage strain on a restrained joint tears the mushy weld open. Look for centerline, crater, or transverse cracks with oxidized, discolored faces.
Three conditions at once: hydrogen in the arc, a hard martensitic HAZ from fast cooling, and restraint stress. Delayed — can appear hours or days after welding. Look for toe or root cracks in the HAZ with clean, bright faces.
The unfilled crater shrinks as it freezes and pulls itself apart — often a star of fine lines. Same mechanism in every process, same cure: fill the crater before stopping. See crater defects for the dedicated reference.
Hot cracking is a solidification story. As the weld metal freezes, the last liquid to solidify is enriched with low-melting constituents — sulfur, phosphorus, and copper in steel; eutectic-forming elements in aluminum. Those films remain liquid at the grain boundaries after the surrounding metal has gone solid. Meanwhile the solidifying weld shrinks, and restraint piles that shrinkage strain onto the still-liquid films. When the strain exceeds the film's ductility, the weld opens along the centerline. The classic recipe: a crack-prone alloy, a high-sulfur or phosphorus base metal, high heat input, and a joint that cannot move.
Cold cracking is a HAZ disease. Hydrogen from moisture dissolves into the molten pool and diffuses into the heat-affected zone, where it waits at trapping sites. A hard martensitic microstructure forms when hardenable steel cools fast — no preheat, cold plate, thin section. Restraint supplies the stress. When the stress and the hardness cross a threshold, the HAZ lets go — which is why a weld can look perfect on Friday and crack by Monday.
Lamellar tearing belongs to none of these families — the weld is sound, and the plate tears under it in terraced steps along inclusions rolled into the steel. Preheat, low-hydrogen rods, and slow cooling do not stop it; it is a design and material problem. See lamellar tearing for the full treatment.
Cracking is the one defect where the cure starts before the steel arrives: material choice, joint design, and procedure. The welder's checklist is short and non-negotiable.
Confirm the steel is weldable and the filler is the right class. Hardenable or high-strength steel? Set preheat and interpass per the WPS and verify with a contact thermometer or temp stick. Moisture at the joint: degrease, dry, and keep the rod oven working — low-hydrogen electrodes live in a rod box, not on the bench.
Stay inside the qualified heat-input band — not too hot (segregation, wide mushy zone), not too cold (martensite). Hold interpass temperature; do not let the weld cool between passes on hardenable steel. Run stringers or a controlled weave and fill every crater before you lift the hood.
On multi-pass joints, watch the HAZ for hairline indications before covering them. Keep each pass within interpass limits, and brush each bead clean so the next pass does not feed contamination to the arc.
The mistake that costs the most: using preheat to "burn out" moisture from a wet joint or damp rods. Hydrogen hiding in the joint still finds its way into the HAZ. Dry, clean, and store consumables properly — then use preheat for what it is for: slowing the cooling rate.
A cracked weld is almost always a reject under the code. If the governing document permits repair, the excavation must go well past the crack tips.
| Step | Action | Why |
|---|---|---|
| 1 | Report and quarantine the weld; photograph and document the crack. | Every crack is a code rejection — you need a record and a verdict before touching it. |
| 2 | Locate the full extent — MT or PT on the surface; UT or RT for internal cracks. | Cracks run farther than they look — the tips hide beyond the visible line. |
| 3 | Excavate to sound metal, past both crack tips, with a generous margin. | Re-welding over a crack tip guarantees the crack returns — deeper. |
| 4 | Verify removal with MT or PT on the excavated groove. | The code requires proof of complete removal before repair welding. |
| 5 | Repair to the qualified WPS — preheat, low-hydrogen consumables, controlled heat input. | The repair must be harder to crack than the original weld. |
Some cracks are not repairable. Codes cap the number and size of repairs, and repeated cracking in the same joint — especially hydrogen cracking in high-strength plate — often means the part is scrapped. Confirm repair eligibility with the engineer and the governing code before cutting.
Every process feeds the arc its own brand of crack risk. Read yours against the list.
Aluminum is the MIG hot-crack story: 5052 and 6061 need the right filler — 5xxx-series filler like 5356, or 4043 on 6xxx — and controlled dilution. Too much base metal in the pool, and centerline cracks follow. On steel, spray transfer with excessive heat on a restrained joint can hot-crack the same way.
Austenitic stainless cracks hot when the filler is mismatched or the root is run fast and light — a wide mushy zone and segregation. On aluminum, arc time without filler is autogenous welding, and autogenous welds crack. Balanced parameters and the right filler beat any technique fix.
The classic split: a cellulosic root (6010/6011) gives penetration but adds hydrogen; the low-hydrogen fill (7018) restores toughness. Cold cracks appear when a high-hydrogen electrode runs on hardenable steel with no preheat. On the second pass, clean the root before covering it.
Multi-pass FCAW on high-strength steel is a hydrogen trap if the wire is damp or the interpass temperature drops. Gas-shielded wires hold hydrogen in check; self-shielded wires tolerate dirt but not moisture in the core. Keep the wire dry and the interpass up.
High heat input and high dilution pull sulfur and phosphorus out of the plate and into the weld — the classic recipe for hot cracking on restrained joints. Single-pass deep welds on high-sulfur steel are the danger zone. Watch flux moisture; it is a hydrogen source like any other.
High current density on aluminum or stainless with fast travel gives the pool no time to dilute properly — hot cracks follow. A worn orifice makes it worse. When it appears, check the consumables first, then the speed.
For hot cracks, fix chemistry, dilution, and heat input. For cold cracks, fix preheat, hydrogen, and restraint — in that order.
| Parameter | Check | Typical Fix |
|---|---|---|
| Preheat / interpass | Measured on the joint per WPS — temp stick or contact thermometer | Hardenable steel: preheat to spec; hold interpass between passes |
| Heat input | kJ/mm or kJ/in from current × voltage ÷ travel speed | Too high feeds hot cracks; too low feeds cold cracks — hold the qualified band |
| Electrode hydrogen class | H4/H8/H16 marking on the electrode; storage and re-dry logs | Specify H4/H8 for hardenable steel; re-dry per the manufacturer |
| Moisture control | Rod oven temperature, covered wire, dry joint | Store low-hydrogen at 250–300°F; re-dry after exposure; dry damp joints |
| Restraint | Joint rigidity, clamp pattern, fit-up gaps, weld sequence | Ease clamps, balance the sequence, let the joint move |
| Cooling rate | Interpass delay, ambient temperature, plate thickness | Hold interpass; slow the cool on hardenable steel; avoid cold drafts |
| Current | Actual output vs the dial; arc stability | Excessive current widens the mushy zone — use the qualified range |
Use the Heat Input Calculator to hold the qualified band, the Filler Metal Selector to pick a low-hydrogen or crack-resistant filler, and the Material Thickness guide for preheat guidance.
The machine matters less than what it feeds the arc — but three links in the chain carry most of the blame.
Power sources: verify the dialed amperage matches the actual output — a drifting machine changes heat input without warning. Consumables: the hydrogen class and storage discipline of electrodes separate a sound weld from a delayed crack. Torches and feeders: any moisture or contamination path — damp liner, leaking torch, damp wire spool — feeds hydrogen straight to the arc.
Cracks are found visually when they are open — and by MT, PT, or UT when they are not. The delayed crack is why timing is part of the inspection.
Cracks are the exception to code tolerance: nearly every governing document — AWS D1.1, ISO 5817, API 1104 — prohibits cracks of any size, whatever their length or location. The rare exceptions are written into the contract, such as specified micro-check cracks in hardfacing overlays. Read the code that governs your job, but assume any crack is a reject.
The questions welders actually ask about cracks — answered straight.
By location, timing, and the fracture faces. Hot cracks are in the weld metal — centerline, crater, transverse — and they appear immediately, with oxidized, discolored faces. Cold (hydrogen) cracks live in the HAZ at the toe or root, appear hours to days later, and have clean, bright faces. Different locations, different timing, different cure.
That is hydrogen cracking. Hydrogen from moisture diffuses into the heat-affected zone and waits at trapping sites while the weld cools to room temperature. When stress and the hard martensitic microstructure cross a threshold — often hours or days later — the HAZ lets go. It is why hardenable steel gets MT at 48 hours.
Often, but never by welding over it. Excavate past both crack tips into sound metal, verify complete removal with MT or PT, then re-weld with preheat and low-hydrogen consumables. Codes cap the number and size of repairs, and repeated cracking — especially in high-strength plate — often means the part is scrapped. Confirm with the engineer and the governing code first.
No. Preheat is the main defense against cold (hydrogen) cracking — it slows cooling, prevents martensite, and dries the joint. It does little for hot cracking, which is a chemistry and dilution problem: sulfur, phosphorus, or a crack-prone alloy will crack at any preheat. Fix hot cracks with filler choice, joint design, and heat input.
Aluminum has high thermal contraction and, in alloys like 5052 and 6061, a wide solidification range with low-melting grain-boundary films. Under restraint, the shrinking weld tears along those films before it finishes freezing. The fix is the right filler alloy and controlled dilution — not more heat.
A crack is a separation in the weld metal or HAZ. Lamellar tearing is the base plate tearing in terraced steps along inclusions rolled into the steel, driven by through-thickness contraction under the weld toe. The weld itself is sound, and preheat or low-hydrogen practice will not stop it. See lamellar tearing for the full treatment.