The empty dish where the weld stops. When the arc cuts off abruptly, the last pool to freeze shrinks, collapses, and can tear open — leaving a shrinkage void, often with star-shaped cracks radiating from the center. Every process has a termination technique that prevents it.
Crater defects are a family of discontinuities that form in the last weld metal to solidify — the crater at the end of the bead. Two distinct members share the name: the shrinkage void (an underfilled dish, sometimes with pipe porosity open at the surface) and the crater crack (a hot crack, usually star-shaped, radiating from the crater center).
They matter because a crater crack is a solidification crack — a sharp stress raiser placed exactly where service loads concentrate, at the end of the weld. And they are entirely preventable: every process has a stop that backfills the crater. A crater at every weld end is not bad luck, it is a termination habit.
Crater cracks, crater voids, crater shrinkage, pipe porosity, star cracks, end-of-weld cracks.
Read the crater before you touch it — the shape of the stop tells you what the termination habit is doing.
One cause, three flavors: the arc stops before the pool has been backfilled. Everything else — current, technique, restraint — just decides how badly it tears.
The crater is the last pool to solidify, and it is the only part of the weld that solidifies with no molten metal feeding it. When the arc stops, the heat source leaves and the pool freezes from the edges inward. Solidification shrinkage — the same force that makes the whole weld contract — pulls the liquid center down. Without filler to backfill, the surface collapses into a dish: the crater.
The crack forms in the last seconds. The crater center is the very last metal to freeze, and it is trapped: the surrounding weld has already solidified and shrunk, so the still-liquid center is pulled in all directions at once. When the shrinkage strain exceeds what the liquid films between growing grains can take, the center tears open. Because the tear follows the weakest path — the boundaries between the columnar dendrites growing in from the crater walls — it radiates outward from the center: the star crack. It is a solidification crack, a hot crack: it forms at freezing temperature, in the weld metal, immediately. No delay, no hydrogen, no hard heat-affected zone.
Shrinkage of the freezing weld pulls the molten center down and in. No filler to backfill means the surface caves into a dish — the void you can see with your eyes.
The last-to-freeze center is pulled apart along the weakest boundaries — the columnar grain boundaries — producing cracks that radiate from the center like spokes.
When shielding ends with the arc, gas (hydrogen from moisture, air) gets trapped in the still-liquid crater — a pipe or cluster of pores open at the surface.
Restraint is the amplifier. On a free-moving fillet, the pool can accommodate shrinkage; on a thick, clamped, high-restraint joint, every degree of shrinkage is locked in, and the crater center takes the whole load. That is why the same termination habit produces nothing on light work and a star crack on heavy structure.
Crater defects are a habit, not a mystery. Decide how every weld ends before you strike the arc.
Know your termination. If the machine has crater-fill, downslope, or pulse tail-out, set it and prove it on scrap. If it doesn't, plan the back-step stop. Keep a rod oven and dry consumables — wet filler feeds crater porosity.
Approach the end deliberately: ease off travel speed so the pool shrinks as you arrive. Don't accelerate at the finish line. TIG: taper current and keep adding filler right up to the stop.
Back-step: stop short, break the arc, move back about 1/2", re-strike on the finished bead, and weld forward over the crater — the crater fills itself. Or use the machine's crater-fill program. Then hold the torch over the crater for post-flow.
The abrupt lift — whipping the torch or electrode away at the end. The pool collapses, the shield disappears, and you hand the inspector a star crack. And don't just re-strike on top later: the crack is already there under the new metal.
When the crater is already cracked, the repair is the same regardless of process — but only after you have fixed the termination habit that made it.
| Step | Action | Why |
|---|---|---|
| 1 | Fix the termination habit first — decide how the next stop will be made. | Re-welding with the same stop reproduces the defect at the same spot. |
| 2 | Locate the extent — VT, then PT or MT on the crater and the stop region. | Star cracks are hairline; the tips run past what the eye sees. |
| 3 | Grind out the full crater to sound metal, plus a margin past the crack tips. | A hidden crack tip re-cracks the repair instantly. |
| 4 | Verify removal — MT or PT on the excavated groove. | Never re-weld over an unconfirmed crater crack. |
| 5 | Re-weld to the WPS with a proper termination — back-step, downslope, or crater-fill. | The repair pass must end in a filled crater, not a new one. |
Never grind only the visible crack and "burn over" the rest. A crater crack is a stress raiser — buried under a repair pass it guarantees a second, deeper rejection. Confirm the repair scope against the governing code before starting.
Each process ends its arc differently — learn your process's stop.
Use the crater-fill program or pulse tail-out if the machine has one — it keeps feeding wire while the arc fades, filling the dish. Without it, back-step. A long stick-out at the stop loses shielding and feeds crater porosity.
The classic crater process: current downslope (2–5 s) plus a filler habit. The filler is what backfills the dish — downslope alone just shrinks the pool without filling it. Hold the torch over the crater for the post-flow.
Back-step is the standard: stop short, break the arc, move back about 1/2" onto the bead, re-strike, and weld forward over the crater. Shorten the arc as you finish — a long arc at the end starves the crater.
Same back-step as stick — many FCAW guns have a trigger to feather the arc at the stop. Self-shielded wire run with long stick-out at the end produces crater porosity on top of the void.
The crater hides under the flux blanket and the slag — star cracks are found after deslagging. Use the current taper on modern heads and always inspect the stop end of every SAW run.
Programmed crater-fill is the whole answer — and the whole risk: if the program ends the weld wrong, every part on the line carries the same crater. Verify the stop on every program change, not just the first part.
The crater is made in the last second of the weld — check the parameters that govern that second.
| Parameter | Check | Typical Fix |
|---|---|---|
| Termination technique | How the weld ends: abrupt lift, back-step, or programmed fill | Back-step or crater-fill on every stop — no exceptions |
| Current downslope (TIG) | Set on the machine; the pool must shrink before the arc cuts | 2–5 s downslope, add filler during ramp-down |
| Crater-fill settings (MIG) | Program active; fill time matches pool size | Enable crater-fill / pulse tail-out; tune fill time and wire feed |
| Post-flow | Gas continues after the arc stops | 1–3 s; longer for stainless and aluminum |
| Travel speed at the end | Accelerating at the stop enlarges the crater | Hold speed — or slow to shrink the pool before stopping |
The Heat Input Calculator sets the pool size and cooling rate — which decides how big the crater will be and how fast it freezes. High heat input on thick section is exactly when the stop matters most.
The machine's crater features are the difference between a good stop and a habitual defect.
Check the power source's crater-fill, downslope, and pulse tail-out features before the job — and verify the post-flow timer works. A TIG foot pedal or torch with a gas valve gives the welder control of the taper; without them, the crater is pure hand technique. Keep filler and electrodes dry — moisture in the last pool is crater porosity waiting to happen.
The crater is the most inspected inch of the weld — inspectors know exactly where the defects live.
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 crater defects — answered straight.
No. A crater crack is a hot crack — it forms in the last weld metal to solidify, at the moment the pool freezes, driven by shrinkage. Hydrogen cracking is a cold crack that appears hours later in the heat-affected zone of hardenable steel. Different location, different timing, different fix.
The crater center is the very last metal to freeze. As the surrounding weld shrinks, the still-liquid center is pulled apart along the weakest boundaries — the columnar grains growing in from the crater walls — so the crack radiates outward from the center like spokes. The bigger the pool and the more abrupt the stop, the more likely it is.
Taper the current down over 2–5 seconds with the downslope control, keep adding filler right up to the stop so the dish stays filled, then hold the torch over the crater for the post-flow. The added filler replaces the shrinkage that would otherwise pull the pool apart.
Yes — grind the whole crater to sound metal and verify with MT or PT before re-welding. The catch is that star cracks run beyond what the eye sees, so grind past the visible crack tips, or the repair re-cracks immediately.
That is crater porosity — the arc stops, shielding ends, and the still-liquid crater traps gas as it freezes into a pipe or a spongy cluster. Use the machine's crater-fill program or a back-step termination, and hold the gun over the crater for a moment after the stop.