A crack is the one defect with no tolerance band. Porosity and undercut get measured against a limit. A crack of any size is rejectable under every major code, so the first job is not measuring it, it’s figuring out which kind you have, because the cause and the fix change completely depending on the type. Three families cover almost everything you’ll find: hot cracks (solidification cracks) that form while the weld is freezing, cold cracks (hydrogen-induced cracks) that show up hours or days later in the heat-affected zone, and lamellar tearing that opens up in the base plate under through-thickness load.
The fastest way to sort them is timing, then location, then orientation. A crack that’s already there when the weld cools is hot. A crack that appears the next morning is cold. A crack that runs down the center of the bead is usually solidification. A crack hugging the fusion line in the HAZ, parallel to it, is usually hydrogen. Get those three observations and you’ve narrowed it down before you reach for any test gear.
What Is a Hot Crack (Solidification Crack)?
Hot cracks form at high temperature, while the weld pool is still solidifying. As the molten metal freezes, it shrinks. If low-melting-point impurities like sulfur and phosphorus are present, they stay liquid after the surrounding metal has frozen, collecting in thin films between the growing grains. Those liquid films can’t carry the shrinkage stress, so the weld tears itself apart along the grain boundaries before it’s fully solid.
You see hot cracks immediately. They’re often visible in the crater or down the centerline as soon as you knock the slag off, sometimes while the weld is still glowing. Under magnification a hot crack follows the grain boundaries (it’s intergranular) and the fracture face often looks oxidized or has a dull, dendritic texture because it cracked while hot enough to oxidize.
Centerline Cracking
The most common hot crack runs straight down the middle of the bead, lengthwise. Two things drive it. First, weld bead shape: a deep, narrow bead (high depth-to-width ratio) freezes from both sidewalls toward the center, pushing impurities into a weak seam right down the middle. Second, the joint pulls the still-solidifying weld apart faster than it can knit together. Centerline cracks are common in the first pass of a deep groove, in fillet welds with too much penetration, and in high-travel-speed MIG where the bead is thin and tall.
Crater Cracking
When you break the arc too fast at the end of a bead, the crater (the pool that’s left) freezes last and shrinks against the surrounding solid metal with nothing to feed it. It tears, usually into a small star-shaped crack. Crater cracks are a launch point for a longer crack to propagate under load, so they aren’t cosmetic. Fill the crater: pause and let the puddle build at the end of the bead, back up over the bead before breaking the arc, or use the crater-fill function or downslope on the machine.
What Makes Hot Cracking Worse
Sulfur and phosphorus content in the base metal is the big lever, which is why free-machining steels and some resulfurized grades are miserable to weld. Heavily restrained joints, deep narrow beads, high travel speed, and high heat input on crack-sensitive material all push it. Certain alloys are inherently sensitive: some aluminum alloys, fully austenitic stainless steels, and nickel alloys like Inconel are prone to solidification cracking, which is why filler choice and bead shape get so much attention there. For aluminum specifically, filler and dilution control are the main defense, covered in the aluminum cracking material guide rather than here.
What Is a Cold Crack (Hydrogen-Induced Cracking)?
Cold cracking, also called hydrogen-induced cracking, hydrogen-assisted cracking, or delayed cracking, forms after the weld has cooled below about 400F (200C), often hours or days later. The delay is the signature. A weld can pass visual inspection, sit overnight, and crack while nobody’s touching it. Hydrogen-induced cracking in the HAZ is the single most consequential cracking mechanism on carbon and low-alloy steel, and the reason preheat and electrode storage exist as disciplines.
Three conditions have to occur together for it to happen, and removing any one prevents the crack:
- Dissolved hydrogen in the weld and HAZ. It comes from moisture in flux coatings, damp or unbaked low-hydrogen electrodes, and from oil, grease, rust, paint, and water on the joint.
- A susceptible (hard) microstructure, typically martensite, formed when a hardenable steel cools too fast through the transformation range.
- Tensile stress, from joint restraint and the residual stress every weld leaves behind.
As the steel cools, hydrogen that was happily dissolved at high temperature gets trapped. It migrates to the hard, stressed HAZ and embrittles it. Given a day or two, the combination of trapped hydrogen and stress cracks metal that was sound when you finished welding.
Where Cold Cracks Show Up
Most hydrogen cracks live in the HAZ, just outside the fusion line, running roughly parallel to it. Toe cracks start at the weld toe and angle down into the HAZ. Underbead cracks run hidden beneath the weld, parallel to the fusion boundary, and are nasty because they don’t break the surface. Root cracks form at the root of the first pass where restraint and stress concentrate. Under magnification a cold crack is usually transgranular (it cuts across grains) and the fracture face is bright and clean because it cracked cold, with no oxidation.
Preventing Cold Cracking
You attack at least one of the three legs:
- Cut the hydrogen. Use low-hydrogen electrodes (the E7018 family, H4 or H8 designators), keep them in a rod oven at the manufacturer’s temperature, and re-bake exposed rods rather than welding with them damp. Clean the joint of oil, rust, paint, and moisture.
- Slow the cooling rate. Preheat keeps the HAZ from quenching itself into martensite and gives hydrogen time to diffuse out before the metal hardens. The right number depends on carbon equivalent, thickness, and restraint, and the AWS D1.1 and ASME B31.3 preheat tables are the starting point. See the preheat temperature guide for the charts and how to measure it, and hold the interpass temperature within range across multi-pass work.
- Reduce restraint and stress. Better joint design, sensible welding sequence, and not boxing a joint into a corner where it can’t move all help. On crack-sensitive jobs, post-weld heat treatment drives off remaining hydrogen and relieves residual stress.
Carbon equivalent (CE) is the shorthand for how hardenable a steel is. The higher the CE, the more readily the HAZ forms martensite, and the more preheat and low-hydrogen practice matter. As a rough orientation, steels under about 0.40 CE are forgiving, and above roughly 0.45 to 0.50 CE you should be treating preheat and hydrogen control as mandatory rather than optional. Confirm the actual requirement against the mill certificate and the governing code, not a rule of thumb.
What Is Lamellar Tearing?
Lamellar tearing is the third family and it’s a base-metal problem, not a weld-metal one. It’s a step-like crack that forms in the base plate, below and parallel to the fusion line, when the joint loads the plate in the through-thickness direction (the short direction, through the rolled thickness). Rolled steel plate contains flattened non-metallic inclusions (mostly manganese sulfides) stretched out in the rolling plane. Weld shrinkage that pulls on the plate’s thickness can decohere those inclusion planes and link them into a terraced crack.
It’s most common in heavy T-joints and corner joints with large welds and high restraint, where shrinkage acts straight through the plate. The defenses: use plate with guaranteed through-thickness ductility (Z-grade steel tested per through-thickness reduction-of-area), redesign the joint so the shrinkage doesn’t pull on the short transverse direction, butter the plate face with weld metal first, and reduce restraint. Low-hydrogen practice helps here too because hydrogen makes lamellar tearing worse.
How to Tell Which Crack You Have
Walk the three observations in order. Timing rules out the most. Then location and orientation point you the rest of the way.
| Clue | Hot Crack (Solidification) | Cold Crack (Hydrogen) | Lamellar Tearing |
|---|---|---|---|
| Timing | Immediate, while or right after the weld freezes | Delayed, hours to days after cooling | Delayed, after cooling under restraint |
| Location | Weld metal (centerline, crater) | HAZ near the fusion line, toe, root, underbead | Base plate, parallel to and below the fusion line |
| Orientation | Often longitudinal down the bead | Parallel to the fusion boundary | Step-like, terraced, in the rolling plane |
| Fracture face | Intergranular, oxidized/dull, dendritic | Transgranular, bright and clean | Stepped, woody appearance |
| Main driver | Sulfur/phosphorus, bead shape, restraint | Hydrogen + hard HAZ + tensile stress | Through-thickness load on dirty plate |
| Primary defense | Clean base metal, fix bead shape, fill craters | Preheat, low-hydrogen rods, reduce restraint | Z-grade plate, joint redesign, buttering |
The crater is a useful tiebreaker. A star crack sitting in the crater is solidification cracking, full stop. A clean crack running through otherwise-sound HAZ that wasn’t there yesterday is hydrogen. A terraced crack stepping through the plate of a heavily loaded T-joint is lamellar tearing.
Confirming and Sizing the Crack
Surface cracks are confirmed with dye penetrant testing, which pulls a colored or fluorescent dye into the crack and bleeds it back out against a developer so even a tight, hairline crack stands out. On ferromagnetic steel, magnetic particle testing finds surface and slightly subsurface cracks. Underbead and other buried cracks need volumetric NDE, ultrasonic or radiographic testing, because nothing on the surface tells you they’re there. The delayed-cracking problem is exactly why code work often requires NDE be held until 24 or 48 hours after welding on certain steels: inspect too soon and the hydrogen crack hasn’t formed yet.
For the surface methods and their limits, see the dye penetrant testing guide and the lack of fusion guide, since lack of fusion and cracks both show as planar reflectors and get confused on the gear by inexperienced operators.
Repairing a Cracked Weld
The wrong move is welding over a crack or grinding the surface until it looks clean. A crack tip is a stress concentrator, and a buried or surface-blended crack tends to reopen and grow under load. The general sequence on any crack that matters:
- Find the ends. Use PT or MT to mark where the crack actually starts and stops. Cracks usually run farther than the visible portion. Drilling a small stop hole just past each end is a common way to keep it from running while you work.
- Excavate to sound metal. Grind or arc-gouge the entire crack out, slightly past both ends and through the full depth. Verify with PT or MT that the cavity walls are crack-free before you put any metal back.
- Find and fix the cause. If it was hydrogen, the repair needs preheat and dry low-hydrogen rods or you’ll just crack it again. If it was solidification, fix the bead shape or clean the base metal. Repeating the original conditions reproduces the original crack.
- Re-weld per the WPS with the corrected practice, then re-inspect.
On structural or pressure work, none of that is freelance. A cracked structural or pressure weld is an engineering matter: the disposition (repair, replace, or accept by analysis) and the repair procedure have to follow the governing code’s repair provisions, AWS D1.1, ASME, or API 1104 as applicable, with sign-off by a qualified inspector and the engineer of record. No repair guarantees a crack won’t recur, which is exactly why the cause has to be corrected and the repair re-inspected rather than trusted on appearance. The hot-work itself follows normal welding safety practice per ANSI Z49.1.
Common Mistakes
Calling every crack a bad rod. The rod is rarely the cause. A centerline crack is bead shape and restraint. A delayed HAZ crack is hydrogen and cooling rate. Diagnose the type before you blame the consumable.
Welding without preheat on hardenable or thick steel. This is the classic setup for a Monday-morning crack. The weld looked perfect Friday because the hydrogen crack hadn’t formed yet.
Storing low-hydrogen electrodes on the bench. E7018 picks up moisture from the air in hours. Damp low-hydrogen rods defeat the entire point of using them. Rod oven, or re-bake before use.
Quitting the bead too fast. A snapped arc leaves an unfilled crater that cracks. Fill it every time.
Grinding a crack flush and moving on. A blended-over crack is still a crack with a sharp tip. It has to be excavated, not hidden.
For the other rejectable defects, see the lack of fusion guide, porosity guide, and undercut guide. For the cooling-rate controls that prevent hydrogen cracking, see the preheat temperature guide and interpass temperature guide. To confirm a suspected crack at the surface, see the dye penetrant testing guide. Return to weld defects or the welding techniques pillar.