Weld metal deposited without ever bonding to the base metal or the previous pass. The most dangerous defect in the book — invisible on the surface, invisible to the eye, and it fails in fatigue.
Lack of fusion is the failure of weld metal to form a metallurgical bond with the base metal or with a previously deposited pass. The weld metal is there — the joint is full, the bead is shaped — but at the fusion boundary the interface is not welded. It is a plane of mechanical contact separated by oxide film, scale, or simply metal that froze before it melted the surface it was laid on.
This is why it is critical: an unbonded interface is a built-in crack-like discontinuity. It carries no load, it acts as a stress concentrator, and under fatigue or impact it behaves like a pre-existing crack in the middle of a weld that looks finished. It is invisible on the surface — found only by radiography, ultrasonic testing, or by cutting the weld open. A weld with lack of fusion can pass visual inspection and fail in service.
Incomplete fusion, cold lap, cold shut, lack of sidewall fusion, lack of inter-run fusion, lack of root fusion.
Most lack of fusion is invisible. What the eye can see is a hint; what NDT shows is the verdict.
Every cause comes back to one failure: the substrate never got hot enough to melt, or something sat between the pool and the metal that needed to be welded.
Fusion is a melting event. Weld metal bonds only when the arc raises the substrate — the base metal or the previous pass — above its melting point and the two liquid pools mix. When that happens you get a real fusion line: grains grow across the interface and the weld and the base become one piece. When it does not happen, the deposited metal freezes against a solid surface. The result looks like a weld, holds like a tack, and behaves like a crack.
Barriers make it easy. Oxide film, mill scale, slag, grease — each acts as a thermal and a metal barrier between the pool and the substrate. The arc either has to melt through the barrier or push it out of the pool before metal can bond. A cold, dirty, or fast-moving arc leaves the barrier in place, and the filler freezes on top of it.
Geometry is the third ingredient. A leading angle that is too steep points the arc down the filler wire instead of into the joint — so the wire melts and the sidewall stays solid. Low heat input and high travel speed finish the job: the pool freezes before the sidewall ever reached melting temperature.
The first pass fails to bond to the bottom of the groove or the root face. The root gap looks filled; the interface with the joint is not fused. The most common type — and the most dangerous, because everything above it hides it.
Fill passes that ride the centerline never touch the groove faces. The weld "floats" inside the joint — full on the outside, attached to nothing on the sides. The classic case of weaving too narrowly or angling away from the sidewall.
The next bead freezes on top of the previous pass without melting its surface. Slag or oxide left on the pass, interpass temperature too low, or just laying a pass on without the heat to re-melt the one beneath. Etching shows the pass boundaries as dark bands.
At the toe of a fillet or cap: weld metal rolls over the base without fusing, leaving a crevice between bead and plate. Driven by a flat bead with no wetting action, low heat, or a travel direction that pushes the pool onto the plate instead of into it.
What makes it critical is what happens next. The unbonded interface does not transfer load; under cyclic or impact loading it concentrates stress like a starter crack. It is invisible on the surface and often invisible to every inspection except RT and UT — which is why codes treat it as one of the most serious discontinuities that can be found in a weld.
Lack of fusion is usually a technique defect before it is a parameter defect. Fix the hand first, then fix the numbers.
Grind to bright metal inside the joint — remove oxide, mill scale, paint, and grease from the faces and the root. Brush every pass before the next one. The barrier between pool and base is the first cause of fusion failure.
Heat input must be enough to melt the sidewall, not just the filler. Check it with the Heat Input Calculator, and size the electrode for the joint — a wire too small for a big groove cannot reach the corners.
Keep the arc aimed at the joint line, not down the wire. A shallow leading angle puts the heat on the work; a steep one melts filler into a floating bead. You are welding with the heat — put it where the bond has to happen.
Hold the pool at the sidewall until it wets and bonds — a short dwell, not a flick. Weave to the edges and pause there; a weave that sweeps through the middle of the groove never fuses the faces.
Remove all slag, dross, and oxides between passes, and keep interpass temperature above the WPS minimum. A cold, slag-covered pass is a surface the next one will refuse to weld to.
Don't confuse "it looks welded" with "it fused". And don't reach for more heat alone — if the arc is aimed wrong, more heat just makes a bigger floating bead. Angle and wash come first.
When lack of fusion is already in the weld, the repair is excavation — every unbonded interface has to come out before anything goes back in.
| Step | Action | Why |
|---|---|---|
| 1 | Fix the cause before touching the weld — angle, heat, cleanliness, or speed. | Re-welding with the same technique reproduces the same unbonded interface. |
| 2 | Map the extent with RT or UT; macro-etch a sample where possible. | Internal lack of fusion extends well beyond anything you can see. |
| 3 | Excavate to sound metal — grind or gouge out every unbonded interface. | The interface is a crack; a repair pass laid over it buries a crack. |
| 4 | Verify removal — VT, PT, or MT on the groove faces. | Never re-weld over an unconfirmed bond line. |
| 5 | Re-weld to the qualified WPS with corrected technique and parameters. | The repair pass must meet the same standard as the original weld. |
Never weld over the top of lack of fusion — the unbonded interface stays, now buried deeper and harder to detect. Excavate fully, confirm the repair scope against the applicable code, and only then re-weld.
Each process fails fusion in its own way. Read yours against the list.
Spray transfer run too cold drops into globular — big droplets, poor wetting, no sidewall wash. Also: excessive gun angle, low voltage for the wire size, and cold short-circuit transfer on a thick section that needs the heat of spray.
Low amperage for the joint, filler fed before the pool has wet the sidewalls, and a torch angle that heats the rod instead of the joint. On stainless and aluminum, oxide on the joint face or on the filler prevents the pool from bonding — clean before every pass.
Arc too long — the heat goes into the arc column instead of the joint. Low amperage and a poor rod angle in a narrow groove; slag left between passes is a recurring cause in multipass welds.
Low voltage for the wire feed rate — the wire piles up without wetting. Long stick-out cools the arc before it reaches the joint, and slag not fully removed between passes gives inter-run fusion failures.
Low current or an over-long arc for the joint; a flux blanket too thin or lifted so the arc runs out from under the flux; damp flux that refuses to carry the arc. On the first pass, a large root face that the arc never burns through.
The keyhole never establishes — too little current for the travel speed, or wrong plasma gas flow — so the front wall of the joint never melts through. Also worn nozzle orifices that spread the arc instead of concentrating it.
When lack of fusion appears, check these in order — technique variables first, because that is where this defect lives.
| Parameter | Check | Typical Fix |
|---|---|---|
| Current / heat input | Too low for the joint — the arc melts filler, not the sidewall | Raise amperage to WPS range; verify heat input with the calculator |
| Travel speed | Pool trails the arc; sidewall never reaches melting | Slow down so the arc dwells long enough to melt the joint face |
| Torch / gun angle | Leading angle too steep — arc heats the filler, not the joint | Flatten the angle toward the work; keep the arc on the joint line |
| Arc length | Too long — heat lost in the arc column (stick, TIG) | Shorten the arc; steady the hand and hold the rod in |
| Weave pattern | Edges skipped or flicked; width beyond the electrode range | Dwell at the sidewalls; keep the weave within 2–3× electrode Ø |
| Preheat | Cold section bleeding heat before the pool can fuse | Preheat per WPS on thick or high-strength sections |
| Interpass temperature | Below the WPS minimum between passes | Hold the minimum so each pass melts the one beneath it |
Use the Heat Input Calculator to confirm whether the arc is actually delivering enough heat to the joint, the Welding Calculator to size current and voltage for the wire, and the Amperage Guide to match the electrode to the section.
The hardware between the machine and the arc decides whether the heat lands where it is supposed to.
A power source that sags under load starves the arc; a worn contact tip makes it wander off the joint line. Size the consumables to the joint — a wire or electrode too large for the groove cannot reach the sidewalls — and replace anything that disturbs a stable, concentrated arc.
Lack of fusion is an internal discontinuity — visual inspection finds only its surface edge. The real look happens with NDT.
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 lack of fusion — answered straight.
They fail at different points. Lack of fusion means the weld metal did not bond to the base metal or a previous pass — a metallurgical failure at an interface. Lack of penetration means the weld never reached the root of the joint — a geometric failure. A weld can have both: it can reach the root and fail to bond, or bond perfectly and never reach the bottom of the groove.
Because lack of fusion is internal. The bead surface can be clean and uniform while the sidewalls underneath are just resting on top of the joint — never fused. The eye cannot see an unbonded interface; radiography and ultrasonics can. A weld that looks perfect on the surface can be rejected the moment it is radiographed.
Most forms, no — sidewall fusion and inter-run fusion are inside the weld. The one exception is cold lap at the toe, where a lip of weld metal rolls over the base metal without fusing; a thin crevice between the bead and the plate is often visible. Everything else needs RT, UT, or a macro-etch to confirm.
Grind the cold-lapped metal out to sound metal, then re-weld with the pool allowed to wash into the joint. The fix is technique: flatten the torch angle toward the work, let the pool wet the toe, and don't travel faster than the pool can bond. Cold lap that is left in place is a crack-like notch at the weld toe.
No. Heat aimed at the filler instead of the joint just makes a bigger bead that still is not welded. The arc has to be pointed at the joint line — not down the electrode or the wire — and the pool has to wash into the corner. Increase heat input, yes, but check angle and technique first.
Usually low amperage for the joint, feeding filler before the pool has wet the sidewalls, a torch angle that puts the heat on the rod instead of the joint, or contamination on the base metal and filler. On stainless and aluminum, oxide film on the joint face is a common hidden cause — the pool floats on it instead of bonding.