What Shielding Gas Is Doing and What Happens When It Stops

Why this matters

Shielding gas is not blowing air away from the weld. It is occupying the space the air would otherwise be in, and the moment that occupation is imperfect the pool starts dissolving nitrogen and oxygen out of the atmosphere at a rate liquid steel is very good at. Coverage fails mechanically almost every time - flow, nozzle, stickout, a draft, a leaking liner - and it fails long before anyone has cause to question the gas mixture. The trap is that the symptom looks identical whether the flow was too low or too high, so a shop that treats "more gas" as the universal fix makes half of its coverage problems worse.

Before opening a cylinder or working around one: shielding gases displace breathing air. Argon and carbon dioxide are both heavier than air and collect in pits, tank bottoms and sumps, which turns a routine space into an oxygen-deficient one with no smell and no warning. Welding in a tank, pit or vessel is a permit space question under 29 CFR 1910.146 with the welding-specific ventilation requirements at 29 CFR 1910.252(c), needing atmospheric testing before entry and continuous ventilation during work, not a fan pointed at the opening. Never use oxygen to ventilate a space or to blow off clothing; oxygen enrichment makes ordinary clothing and oily rags burn violently, and ANSI/AWS Z49.1, in the edition your employer's written program adopts, prohibits it outright. Secure cylinders upright with the valve protection cap on when not in use under 29 CFR 1910.253. The arc's ultraviolet converts surrounding oxygen to ozone, most strongly under argon-rich shielding on aluminium and stainless, and carbon-dioxide shielding decomposes in the arc to carbon monoxide; both carry limits in 29 CFR 1910.1000 Table Z-1 and both are inhalation routes needing local exhaust at the arc or a respirator under a written program meeting 29 CFR 1910.134.

What the gas is holding out

Air is roughly 78 percent nitrogen and 21 percent oxygen, and molten steel dissolves both readily. Two things then happen.

Oxygen burns the deoxidisers out of the deposit. Silicon and manganese in the filler are there partly to scavenge oxygen. Feed the pool atmospheric oxygen and they are consumed doing that instead of alloying, the deposit loses toughness, and the surplus shows up as glassy silicate islands on the bead face.

Nitrogen dissolves happily in the liquid and very poorly in the solid. As the pool freezes, the solubility collapses and the nitrogen has to go somewhere: out through the surface if it has time, into gas pockets if it does not, or into hard nitride particles that reduce ductility. Which of those you get is a solidification-rate question, and reading the resulting pore pattern is owned by the porosity card in this library.

The gas therefore has one job with two ends: cover the pool from arc initiation until the metal has cooled below the temperature where it will still react. That is why post-flow exists on gas tungsten arc welding and why a welder who whips the torch away at the crater gets a discoloured, sometimes cracked crater on an otherwise clean bead.

Coverage is a mechanical problem before it is a chemical one

In field practice the order of suspicion runs like this, and the mixture is near the bottom of it.

• Flow rate outside the working band for the nozzle in use. A common band for gas metal arc with a standard nozzle sits roughly between 20 and 40 cubic feet per hour, and that band is stated at constant nozzle size, ordinary stickout and no draft - a larger cup, a longer stickout or an open door all move it. Your consumable supplier and your welding procedure specification own the number for your setup.
• Spatter build-up in the nozzle. A partially bridged nozzle both restricts flow and turns what gets through into a jet, which is the worst of both errors at once.
• Excessive stickout or nozzle-to-work distance. The gas column spreads as it leaves the cup, so the further the cup is from the pool, the thinner and wider the cover.
• Draft at the joint. Gas-shielded processes are commonly derated above roughly 5 miles per hour of air movement at the joint, held at constant nozzle size and standard flow; the site-selection consequences of that number belong to the arc-processes card.
• Leaks upstream of the gun. A cracked hose, a loose fitting or a worn liner O-ring lets gas out and, on the suction side of some feeders, lets air in. A leak that whistles is easy; a leak that does not is found by a soap solution on every joint from regulator to gun.
• Joint geometry that will not hold gas. An outside corner, a lap with a wide gap, an overhead joint where the gas falls away from the pool. Geometry can defeat a perfectly set flow.
• Then, and only then, the mixture.

What the mixture actually buys

Composition changes arc behaviour and bead shape more than it changes protection, which is why it is the last thing to suspect for a coverage failure and the first thing to adjust for a profile problem.

• Pure argon is fully inert and gives a soft arc, but on steel it produces poor wetting and an unstable arc, which is why it is used straight on aluminium and non-ferrous work and almost never straight on carbon steel with a solid wire.
• Carbon dioxide additions make the arc hotter and more penetrating and widen the fusion zone at the root, at the cost of more spatter and a rougher bead. Straight carbon dioxide gives the deepest penetration of the common choices and will not support spray transfer.
• Small oxygen additions stabilise the arc and improve wetting at the toes on steel, which reduces one of the two toe defects the undercut and overlap card covers.
• Helium additions raise the arc voltage at a given length and put more energy into the pool, which helps on thick sections and high-conductivity metals. Helium is much lighter than air, so it needs higher flow to hold coverage at the same joint.
• Back-purging is a separate question from torch-side shielding. Where the root side of stainless or a similar alloy is exposed to atmosphere while hot, an unpurged root oxidises into a rough grey or black scale that is a corrosion initiation site, and no amount of torch-side gas reaches it.

One gate, two jobs that failed it in opposite directions

The gate: the pool must sit under a continuous, low-turbulence column of gas from arc start to well past arc stop. Both jobs below broke that gate. The symptom on the bead face was the same. The corrections were opposite, and each shop's instinct would have made the other one worse.

Job A: a bench weld on a 1/4 in carbon steel bracket, indoors, doors shut. Gas metal arc, solid wire, standard nozzle. The welder set flow at 15 cubic feet per hour, below the 20 to 40 band stated above for that nozzle at ordinary stickout. Halfway through a run of brackets the deposits started showing surface pinholes. The welder shortened the stickout, which helped for two parts and then stopped helping.

The flowmeter still read 15 with the trigger pulled, and the nozzle had a ring of spatter reducing the opening by roughly a third of its diameter. Two faults compounding: a set point below the band, and a restriction cutting what reached the pool below even that.

Correction, printed: a flowmeter reading is not a delivery figure. The 15 on the gauge is flow through the regulator, and it already contains whatever restriction sits downstream of it as a pressure the gauge cannot show. Re-basing rather than adding: the action was not "add 10 to the 15" but "clear the restriction, re-set the flow into the band, and re-read it with the trigger pulled." The shop cleaned the nozzle and set 25 cubic feet per hour, mid-band, and the pinholes stopped on the next part.

Job B: a fillet in an inside corner of a fabricated box, same shop, same machine, same week. The welder had heard about Job A. The joint was tucked into a corner where he could not see the pool well, so he set flow at 60 cubic feet per hour on the reasoning that more gas cannot hurt. The bead came out with the same surface pinholes.

Above the band, the gas column leaving the cup goes turbulent, and a turbulent jet mixes with surrounding air and drags it into the shielding envelope. At 60 against a 40 top of band, this was 1.5 times the upper limit, and the aspiration was doing more damage than the corner geometry ever would have. The fix was to bring flow down to 30, mid-band, and fit a gas lens, which restores a laminar column and buys back the stickout the corner was costing.

Check against the sibling rules, with the figures printed. The arc-processes card states the roughly 5 mph derate point for gas shielding; both of these jobs were indoors with doors shut and measured no meaningful air movement, so neither is attributed to draft, and the 5 mph line is not invoked to explain either. The porosity card owns the reading of pore shape and distribution; this card reports only that surface pinholes appeared and does not diagnose the gas species from their shape. The 20 to 40 band is stated once in the general section and both cases are judged against that same band: Job A at 15 is below it and Job B at 60 is 1.5 times its top, and the corrected settings of 25 and 30 both sit inside it.

What each shop would have concluded alone. Job A's welder, working from Job B's rule, would have run flow up and made a restricted nozzle into a restricted turbulent nozzle. Job B's welder, working from Job A's rule, did exactly that and got there by himself. The symptom carries no direction information, which is the reason to check the delivered flow and the nozzle condition against a stated band rather than reasoning from the bead.

How to verify coverage before you blame the wire

• Read flow with the trigger pulled, not at rest. A static reading describes the regulator, not the gas reaching the pool.
• Look into the nozzle every time you change parts, and clear spatter with the gun de-energised at the feeder rather than by triggering it against a surface. Triggering a gun to clear a bridged nozzle strikes an arc at an uncontrolled point, which is how a shop gets an arc strike on a finished face.
• Soap-test every gas connection from regulator to gun when a coverage problem survives the obvious checks. A slow leak is invisible, bleeds gas continuously, and reads as an intermittent quality problem because it interacts with hose position.
• Confirm post-flow before deciding the crater is a technique fault. A discoloured or cracked crater on a clean bead is a coverage-at-the-end problem, not a travel-speed problem. On gas tungsten arc, an old and serviceable rule is about one second of post-flow per ten amps of welding current, stated at constant cup size and with no draft, with the tungsten held over the crater for that whole time.
• Prove the purge separately where the root matters. Torch-side gas says nothing about root-side atmosphere. If the root is oxidising, the purge failed, and the fix is on the other side of the joint.

References

• 29 CFR 1910.146 for permit-required confined spaces and 29 CFR 1910.252(c) for ventilation and protection in welding, cutting and heating, including work in confined spaces
• 29 CFR 1910.253 for compressed gas cylinder storage and handling; 29 CFR 1910.1000 Table Z-1 for ozone and carbon monoxide; 29 CFR 1910.134 for respiratory protection
• ANSI/AWS Z49.1, Safety in Welding, Cutting, and Allied Processes, in the edition your employer's written safety program adopts, including its prohibition on using oxygen for ventilation
• Consumable and gas supplier documentation for the flow band matching your nozzle, wire and joint, and the welding procedure specification that governs the work
• See related: Why Porosity Appears and What It Tells You; The Common Arc Processes and What Each One Buys; How Undercut and Overlap Each Form