Friction Stir Welding

Welding that never melts the metal

Every welding process you learned about in school — stick, MIG, TIG, oxy-acetylene, laser, electron beam — works by melting. You pour heat into a joint until the metal turns liquid, the two edges fuse into a shared pool, and the pool freezes into a single piece. It is intuitive, it is universal, and on high-strength aluminum it is a quiet disaster. Aluminum loves to dissolve hydrogen when it is molten and spit it back out as porosity when it freezes; the heat ruins the carefully engineered temper of the surrounding metal; and the re-solidified weld bead is a coarse cast structure far weaker than the rolled plate it joined.

Friction stir welding throws the melting away entirely. Invented at The Welding Institute (TWI) in Cambridge in 1991, it is a solid-state process: the metal is heated only to the point where it becomes soft and plastic — think modelling clay or warm toffee — and then physically stirred together. A rotating tool, consisting of a broad flat shoulder and a shorter profiled pin, is plunged into the joint line between two firmly clamped plates and then traversed along the seam. Friction between the shoulder and the surface, combined with the work of shearing the metal, generates the heat. The pin sweeps softened material from the front of the tool around to the back, where the trailing shoulder forges it into a dense, fine-grained weld.

The result is a joint that looks, under the microscope, more like forged metal than welded metal. There is no fusion zone, no cast dendrites, no gas porosity, and very little distortion. For aluminum airframe panels, rocket propellant tanks, and railway carriage bodies, that combination is so valuable that friction stir welding has become the default rather than the exception.

The mechanism: shoulder, pin, and four zones

The tool does two jobs at once. The shoulder — typically 12 to 25 mm in diameter for aluminum plate — rubs flat against the top surface and produces most of the frictional heat, while also containing the plasticized metal so it cannot squirt out as flash. The pin (or probe) is shorter than the plate is thick, usually 95 percent of the plate thickness, and is threaded or fluted so that as it spins it drives material downward and around, mechanically mixing the two sides.

Because the tool both rotates and translates, the two sides of the weld are not symmetric. On the advancing side the tool's surface velocity points the same way as the travel direction; on the retreating side they oppose. Material flows preferentially from the retreating side around to the advancing side, which is why defects almost always show up on the advancing side. A transverse slice through a finished weld reveals four distinct regions:

• Stir zone (nugget / weld nugget). The metal that passed directly around the pin. Dynamic recrystallization here refines the grain to 1–10 microns — finer than the parent plate — giving the nugget excellent strength and ductility. You can often see onion-ring banding from the cyclic deposition of material once per tool revolution.
• Thermo-mechanically affected zone (TMAZ). A band just outside the nugget that was both heated and plastically deformed, but not enough to recrystallize. The grains are bent and elongated, swept up toward the surface.
• Heat-affected zone (HAZ). Heated but not mechanically worked. In age-hardened alloys this is the weakest spot in the whole joint, because the thermal cycle over-ages the precipitates and softens the metal — exactly the same problem an arc weld has, but milder.
• Parent metal. Far enough from the seam that it never got hot enough to change.

The governing physics — how much heat, and where

The single most important process parameter is heat input per unit length, and it is set by the ratio of rotation speed to travel speed. The total power delivered to the weld is, to good approximation, the torque the tool resists times its angular velocity:

P = T · ω
  T  = tool torque (N·m), resisting the rotating shear
  ω  = 2π · (rpm) / 60   (rad/s)

The heat input per millimetre of weld is that power divided by the travel speed, and the peak temperature scales with the ratio rpm²/v (rotation squared over traverse). A useful design relation expresses the homologous temperature — the fraction of the absolute melting point the metal reaches:

T_peak / T_melt  ≈  K · ( ω² / v )^α

Typical aluminum FSW:
  T_peak / T_melt  ≈  0.7 – 0.9   (absolute / Kelvin)
  T_peak           ≈  450 – 550 °C   (for 6061, T_melt ≈ 652 °C)
  α                ≈  0.04 – 0.06   (weak power-law sensitivity)

That homologous window — about 0.8 of the absolute melting temperature — is the sweet spot. Below roughly 0.7, the metal is too stiff to flow and consolidate, and you get voids. Above roughly 0.9 you risk incipient melting at grain boundaries and badly coarsened, over-aged metal. Let us size a real weld:

Workpiece:   6 mm 6061-T6 aluminium plate butt joint
Tool:         shoulder 18 mm, pin 6 mm dia × 5.7 mm long, threaded
Parameters:   ω = 800 rpm  →  83.8 rad/s
              v = 150 mm/min  =  2.5 mm/s
              measured torque T ≈ 25 N·m, forge force ≈ 9 kN

Power:        P = T·ω = 25 × 83.8 ≈ 2.1 kW
Heat input:   Q = P / v = 2100 W / 2.5 mm/s ≈ 840 J/mm
Peak temp:    ≈ 500 °C   (773 K / 925 K ≈ 0.84 of T_melt = 652 °C)

For comparison, a MIG weld on the same plate dumps
~1,500–2,500 J/mm and locally reaches 1,200 °C+ (molten).

Notice that FSW achieves a sound weld at roughly half the heat input of arc welding and never goes liquid. That low, contained heat input is exactly why the distortion is small and the heat-affected zone narrow: less energy spread over less width.

Forces, machines, and the forge load

Friction stir welding is as much a forging operation as a welding one, and the forces are large. The dominant load is the downward forge force that keeps the shoulder pressed against the plate and consolidates the metal behind the pin. For 6 mm aluminum it runs 8–15 kN; for 25 mm thick plate it can exceed 50 kN; for steel, well over 100 kN. There is also a sideways traverse force the machine must overcome to push the tool along, and a spindle torque of tens of newton-metres.

This is why early FSW was done on heavily modified milling machines and why purpose-built FSW machines have stiff gantries and high-thrust spindles. Modern systems run the tool on a robot or, more commonly, on a dedicated machine with force control: rather than commanding a fixed plunge depth, the controller commands a fixed forge force and lets the tool find its own depth, which keeps the weld consistent even when the plate thickness or fixturing varies slightly. The plates themselves must be clamped hard against a rigid backing anvil, because the forge force would otherwise simply push them apart.

Friction stir welding versus fusion arc welding

The cleanest way to understand FSW's appeal is to put it side by side with the fusion process it most often replaces on aluminum.

Property	Friction stir welding	MIG / TIG arc welding	Laser beam welding
State of metal	Solid (plasticized)	Molten pool	Molten keyhole
Peak temperature	450–550 °C (no melt)	1,200 °C+ (molten)	2,000 °C+ (vaporizing)
Heat input	500–1,000 J/mm	1,500–2,500 J/mm	200–600 J/mm
Filler / shielding gas	None	Both required	Optional
Porosity	None (no liquid)	Common (hydrogen)	Possible
Joint efficiency	70–90 %	50–70 %	60–80 %
Distortion	Very low	High	Low–medium
Weldable 2xxx / 7xxx Al	Yes	No (hot-cracks)	Marginal
Position freedom	Mostly flat / linear	All positions	Mostly flat
Exit hole at weld end	Yes (or run-off tab)	No	No

The two columns that justify the whole technology are "weldable 2xxx / 7xxx aluminum" and "joint efficiency." The high-strength aerospace alloys 2024 and 7075 are effectively unweldable by arc — they hot-crack in the molten pool — yet they friction-stir-weld cleanly. And because the stir zone is wrought rather than cast, FSW keeps far more of the parent metal's strength than any fusion process can.

Failure modes and the trade-offs

FSW is not magic; it has a characteristic defect catalogue, and almost every defect traces back to getting the heat or the flow wrong.

• Wormhole / tunnel void. Too little heat (low rpm, fast travel, or too little forge force) leaves the metal too stiff to consolidate, and a continuous tunnel of voids runs along the advancing side just below the surface. The single most common production defect. Cure: more rotation, slower travel, or higher forge force.
• Lack of penetration / kissing bond. If the pin is too short, the bottom of the joint is stirred but not fully bonded — the two faces touch but do not metallurgically join. It is dangerous because the touching surfaces can hide from ultrasonic inspection while carrying almost no load. Cure: pin length set to 95–98 percent of plate thickness, and root-pass verification.
• Excessive flash and surface galling. Too much heat over-softens the metal, which extrudes from under the shoulder as flash and thins the weld. Cure: less rotation, faster travel, or a scrolled shoulder that pushes material inward.
• HAZ softening. In age-hardened alloys the heat-affected zone over-ages and becomes the weakest point of the joint — a tensile coupon will neck and fail in the HAZ, not the nugget. This is the fundamental limit on joint efficiency. Cure: post-weld heat treatment, or designing the joint so the HAZ sees lower stress.
• The keyhole exit. When the tool retracts at the end of a weld it leaves a hole the size of the pin. In conventional FSW this is placed on a sacrificial run-off tab and machined away. Retractable-pin tools and friction-stir-spot variants exist precisely to eliminate it.
• Tool wear. A non-issue for aluminum but the dominant cost for steel and titanium, where the tool runs above 1,000 °C and must be made of polycrystalline cubic boron nitride (PCBN) or tungsten-rhenium. A single steel-welding tool can cost thousands of dollars and last only a few metres of weld.

The trade-offs against fusion welding are real: FSW needs heavy fixturing and a stiff machine because of the forge force; it is best at long, straight, accessible seams and awkward at complex 3-D geometry; it leaves an exit hole; and it cannot weld out of position the way a skilled TIG welder can lay a bead overhead. The decision rule is simple — if you have long aluminum seams that demand high integrity and low distortion, FSW wins; if you have short, varied, or positionally awkward joints, arc welding still rules.

Variants of the process

• Friction stir spot welding (FSSW). Plunge, stir, retract — no traverse. A spot-weld replacement for riveting in automotive aluminum bodies. Mazda was an early adopter for aluminum rear doors.
• Self-reacting (bobbin) tool FSW. Two shoulders, one above and one below the plate, joined by the pin, so the forge force is reacted internally and no backing anvil is needed. Used for closed hollow extrusions where you cannot put an anvil behind the weld.
• Retractable-pin FSW. The pin withdraws as the weld finishes, closing the keyhole. NASA developed it for the Space Shuttle external tank circumferential welds.
• Friction stir processing (FSP). The same tool used not to join but to locally refine the microstructure of a single plate — homogenizing cast surfaces, healing porosity, or creating surface composites.
• Stationary-shoulder FSW. The shoulder does not rotate, only the pin does, giving a smoother surface finish and lower heat — favoured for titanium.

Where friction stir welding actually shows up

• Space launch. SpaceX friction-stir-welds the aluminum-lithium (Al-Li 2195) propellant tanks of the Falcon 9. NASA used FSW on the Space Shuttle external tank and on the SLS core stage — barrels of aluminum-lithium up to 8.4 m in diameter, welded with retractable-pin and self-reacting tools.
• Commercial aerospace. Eclipse Aviation built the first FSW-certified business jet airframe, replacing tens of thousands of rivets with continuous welds. Airbus and Embraer use FSW on fuselage and floor panels.
• Rail. Hitachi's A-Train and the bodies of many high-speed trains are built from long extruded aluminum hollow profiles friction-stir-welded into single panels several metres long — light, stiff, and distortion-free.
• Shipbuilding. Aluminum deck and superstructure panels for ferries and naval craft, prefabricated as large FSW panels and then assembled.
• Automotive and electronics. Battery trays and enclosures for electric vehicles, tailgates, bumper beams, and the sealed lids of EV battery packs. Apple has friction-stir-welded iMac enclosures. The leak-tight, fine-grained seam is ideal for sealing liquid-cooled battery cold plates.

Common pitfalls when specifying FSW

• Underestimating fixturing. The forge force will lift, spread, or shift inadequately clamped plates. Design the fixture and backing anvil first; they are not an afterthought.
• Ignoring the advancing/retreating asymmetry. Defects favour the advancing side. Place inspection and the most highly stressed material accordingly, and remember that reversing the weld direction moves the asymmetry.
• Forgetting the HAZ. The weld will fail in the over-aged heat-affected zone of an age-hardened alloy, not in the strong nugget. Size the joint for HAZ strength, and consider post-weld ageing.
• Setting depth instead of force. Position control fights every thickness and flatness variation; force control lets the tool track them. Use force control for production.
• Specifying steel without budgeting for tools. FSW of aluminum is cheap and the tools last; FSW of steel or titanium needs PCBN tooling that wears out fast. The economics are completely different.