Friction stir welding of aluminium alloys

¹TWI, Granta Park, Great Abington CB21 6AL, UK
²BP International, Compass Point, 79-87 Kingston Rd, Staines, Middx TW18 1DY, UK
³Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, UK
⁴School of Materials, University of Manchester, Grosvenor Street, Manchester M1 7HS, UK

*Corresponding author, email philip.withers@man.ac.uk
^† Other terms have been used in the literature, namely 'shear side' and 'flow side', but these are ambiguous and have been discouraged.^[4]

Paper presented at International Materials Reviews, vol.54. no.2. March 2009. pp. 49-93.

The comprehensive body of knowledge that has built up with respect to the friction stir welding (FSW) of aluminium alloys since the technique was invented in 1991 is reviewed. The basic principles of FSW are described, including thermal history and metal flow, before discussing how process parameters affect the weld microstructure and the likelihood of entraining defects. After introducing the characteristic macroscopic features, the microstructural development and related distribution of hardness are reviewed in some detail for the two classes of wrought aluminium alloy (non-heat-treatable and heat-treatable). Finally, the range of mechanical properties that can be achieved is discussed, including consideration of residual stress, fracture, fatigue and corrosion. It is demonstrated that FSW of aluminium is becoming an increasingly mature technology with numerous commercial applications. In spite of this, much remains to be learned about the process and opportunities for further research and development are identified.

Keywords: Friction stir welding, Aluminium alloys, Microstructure evolution, Plastic flow, Residual stress, Mechanical properties, Thermomechanically affected zone.

Introduction

Historical background and principles

Radically new joining processes do not come along very often: friction stir welding (FSW) was one such event, being invented by the TWI in 1991.^[1,2] Since then research and development in FSW and associated technologies has mushroomed, with many companies, research institutes and universities investing heavily in the process and international conference series dedicated to it's study. By the end of 2007, TWI had issued 200 licences for use of the process, and 1900 patent applications had been filed relating to FSW.^[3] The number of research papers has also grown exponentially.

In essence, FSW is very simple, although a brief consideration of the process reveals many subtleties. The principal features are shown in Fig.1. A rotating tool is pressed against the surface of two abutting or overlapping plates. The side of the weld for which the rotating tool moves in the same direction as the traversing direction, is commonly known as the 'advancing side'; the other side, where tool rotation opposes the traversing direction, is known as the 'retreating side'^† . An important feature of the tool is a probe (pin) which protrudes from the base of the tool (the shoulder), and is of a length only marginally less than the thickness of the plate. Frictional heat is generated, principally due to the high normal pressure and shearing action of the shoulder. Friction stir welding can be thought of as a process of constrained extrusion under the action of the tool. The frictional heating causes a softened zone of material to form around the probe. This softened material cannot escape as it is constrained by the tool shoulder. As the tool is traversed along the joint line, material is swept around the tool probe between the retreating side of the tool (where the local motion due to rotation opposes the forward motion) and the surrounding undeformed material. The extruded material is deposited to form a solid phase joint behind the tool. The process is by definition asymmetrical, as most of the deformed material is extruded past the retreating side of the tool. The process generates very high strains and strain rates, both of which are substantially higher than found in other solid state metalworking processes (extrusion, rolling, forging, etc.).

Fig.1. Schematic diagram of FSW process

Friction stir welding is therefore both a deformation and a thermal process, even though there is no bulk fusion. The maximum temperature reached is a matter of some debate. Thermocouple measurements during FSW of aluminium alloys suggest that, in general, the temperature stays below 500°C.^[5-7] These values must be treated with some care, as the position of the thermocouple in the rapidly moving nugget can be difficult to ascertain. Microstructural evidence seems to corroborate the thermocouple based conclusion that unless extreme processing parameters are chosen, the maximum temperature usually lies between 425 and 500°C.^[6,8] It has been suggested that the temperature of the material in contact with the pin may reach the solidus temperature,^[9] although experimental validation is difficult due to the intense deformation at the interface. There is evidence of incipient melting for some aluminium alloys (e.g. 7010) for fast weld speeds.^[10-12] It can also be argued that the peak temperature is inherently self-limiting. The workpiece flow stress will fall rapidly as the solidus is approached, so that heating of the nugget at the tool/workpiece interface limits the available heat generation by reducing the torque.^[13,14]

To date, the predominant focus of FSW has been for welding aluminium alloys, although the process has been well developed for both copper alloys^[15-21] and magnesium alloys.^[22-29] Work is under way to develop the process for materials such as titanium alloys,^[30-34] steels,^[35-43] nickel alloys^[44-46] and even molybdenum.^[47] The welding process in these materials takes place at considerably higher temperatures, and although the feasibility of the process has been demonstrated, further work is needed to improve the performance and longevity of tool materials. In addition considerable work has focused on using FSW to join dissimilar aluminium alloys.^[48-61] Furthermore the steady push to lightweight vehicles has largely been responsible for research in joining aluminium alloys to other metals, including aluminium to magnesium,^[62-65] aluminium to metal matrix composites,^[66] aluminium to steel^[67-69] and aluminium to copper.^[70-73]

Coverage of the present review is confined to the FSW of aluminium alloys. A summary of the AWS designations for wrought Al alloy groups and AWS basic temper designations applicable to heat-treatable Al alloys is contained within Table 1. Since FSW is a solid state process, it can be used to join all common aluminium alloys, including the 2xxx, 7xxx and 8xxx series which are normally challenging or impractical to weld by fusion processes. A key distinction is between non-heat-treatable and heat-treatable alloy series. In work hardened alloys (e.g. 5xxx), the heat from the friction welding process will allow thermal recovery and recrystallisation of dislocation substructures, although this is partly countered in the intensely deformed region where new dislocation structures are generated. In age hardened alloys, the weld will normally be heated well above the dissolution temperature of the initial precipitates, enabling dissolution, reprecipitation and overaging to occur. Friction stir welded aluminium alloys can therefore contain microstructures covering the entire spectrum of normal tempers.

Table 1 AWS designations for wrought Al alloy groups and basic temper designations applicable to heat-treatable Al alloys 

Since its inception, many papers and articles have been published on FSW of aluminium alloys, many of them dealing with microstructure and properties. Recently there have been excellent general reviews of FSW covering a wide range of materials by Mishra and Ma,^[74] which also includes friction stir processing, and by Nandan et al.,^[75] which concentrates on the heat generation, heat transfer and tool/material flow interactions of FSW. A recent ASM speciality handbook also covers FSW and friction stir processing.^[76] Nevertheless, no critical and comprehensive review focusing specifically on aluminium FSW is available in the public domain. It is therefore considered timely to correct this omission. The present review draws on a wide selection of published data to summarise current understanding of the complex relationship between welding parameters, microstructure and properties for FSW of many aluminium alloys. Process modelling of FSW has evolved in parallel with empirical process development, and provides physical insight into all of these relationships. Since FSW modelling has been reviewed elsewhere,^[77,78] this aspect is not explicitly covered in the present review, except where modelling helps to interpret and complement the experimental observations, or to clarify issues debated in the literature.

Advantages/disadvantages of FSW for aluminium joining

The advantages of FSW for welding aluminium can be summarised as follows:

1. as a solid state process it can be applied to all the major aluminium alloys and avoids problems of hot cracking, porosity, element loss, etc. common to aluminium fusion welding processes
2. as a mechanised process (Fig.2a), FSW does not rely on specialised welding skills; indeed manual intervention is seldom required
3. no shielding gas or filler wire is required for aluminium alloys
4. the process is remarkably tolerant to poor quality edge preparation: gaps of up to 20% of plate thickness can be tolerated, although this leads inevitably to a reduction in local section thickness since no filler is added
5. the absence of fusion removes much of the thermal contraction associated with solidification and cooling, leading to significant reductions in distortion; however, it is not a zero distortion technique^[79]
6. it is very flexible, being applied to joining in one, two and three dimensions, being applicable to butt, lap and spot weld geometries; welding can be conducted in any position
7. excellent mechanical properties, competing strongly with welds made by other processes (see the section on 'Comparison with other joining processes')
8. workplace friendly: there are no ulraviolet or electromagnetic radiation hazards as the absence of an arc removes these hazards from the process; the process is no noisier than a milling machine of similar power, and generates virtually zero spatter, fume and other pollutants
9. the energy required at the weld for FSW lies between laser welding (which requires less energy) and metal inert gas (MIG) welding (which typically needs more)*
10. high welding speeds and joint completion rates: in single pass welds in thinner materials (down to 0?5 mm thickness), FSW competes on reasonable terms with fusion processes in terms of welding speed; in thicker materials, FSW can be accomplished in a single pass (e.g. the 50 mm tool in Fig.2d), whereas other processes need multiple passes. This leads to higher joint completion rates for FSW, even though the welding speeds may be lower. Thick plates can also be joined by FSW on either side^[43,80]
11. various mechanical and thermal tensioning strategies can be applied during welding to engineer the state of residual stress in the weld (see the section on 'Residual stress control').

*Note that a distinction is needed between energy required to make the weld and the total energy required to operate the process. The latter depends very much on the specific equipment used, and comparisons are difficult. However, it would be expected that the total energy requirement for FSW would be greater than MIG, but less than laser for single pass welds of the same thickness.

There are of course disadvantages to FSW; indeed, some of the advantages listed above can be viewed in a less positive light in certain circumstances. For example, the absence of a filler wire means that the process cannot easily be used for making fillet welds. Similarly, the fully mechanised nature of the process prevents its use for applications where access or complex weld shape is best suited to a manual process. The presence of a hole at the end of the weld from which the probe was withdrawn is often quoted as a disadvantage. In practice, this has seldom been a significant problem, as there are many possible solutions, which have been considered elsewhere.^[81,82] The workpiece also needs to be restrained in well designed support tooling (Fig.2a), both to react to the forces applied, and to prevent the probe from pushing the workpiece materials apart. Although the process may reduce the strength of aluminium alloys, this can be compensated for if necessary by appropriate design of the joint, for example by locally increasing the thickness, but in most cases no changes are made. Process economics are generally considered favourable, but specific published data are lacking. However, it is known that the process drastically reduces weld preparation costs, skilled welder requirements and repair rates. Efficient power consumption is dependent on matching the size of machine being used to the size of weld being made, although this is not always a practical option.

Fig.2. a) commercial scale FSW machine designed to weld underground train bodies. Workpiece is held by two hydraulic clamps (one is obscured) between which welding head passes from right to left; head has just completed furthermost of three FSWs shown;

b) typical welding head as it is removed from the plate;

c) 25mm shoulder (8mm pin) diameter Triflute MX tool - it is possible to use this tool with zero tilt angle;

d) a 50mm shoulder diameter (50mm pin) Triflat tool designed specifically for welding thicker sections

Applications

Commercial applications have been reported across many industries, and some selected examples are shown below which illustrate the widening appeal of the process. This list is representative rather than exhaustive, and it should be emphasised that new applications are appearing all the time. It should be noted that FSW does not restrict the operating temperature range of aluminium alloys, with applications ranging from cryogenic temperatures (e.g. liquid oxygen and liquid hydrogen rocket fuel tanks) to mildly elevated temperatures (e.g. heat exchangers in heating systems). Most FSWs used in production are butt welds, although lap welds and friction stir spot welds are also being applied with increasing frequency.

Marine

It is believed that the first commercial application of FSW was the joining of 6xxx series alloy extrusions for use in fish freezing plants for fishing vessels.^[83] There have been numerous applications of the process for joining 6xxx extrusions for incorporation in bulkheads and decks in various high speed aluminium vessels, and in large steel cruise ships which now often have lightweight aluminium superstructures.^[84] In such applications, the FSW panels are very flat due to the low distortion, and are cut up and welded into larger structures, usually by MIG welding. Friction stir welding has been used extensively in the aluminium superstructures of cruise ships such as the 'Seven Seas Navigator' which contain many kilometres of friction stir welds, mostly in 6xxx grade extrusions. The world's largest aluminium vessel, the Japanese fast ferry 'Ogasawara', launched in 2004, makes extensive use of FSW in its superstructure.^[85]

Aerospace

The first major application was the use of the process to replace fusion welding in fuel tanks for unmanned Delta II and later Delta IV rockets.^[86-88] The manufacturer (Boeing) has reported virtually zero defect incidence, and significant cost savings over the previous variable polarity plasma arc (VPPA) process. The process has also been adopted for the large fuel tank for the Space Shuttle.^[89-91]

Almost all the major airframe manufacturers are investigating the use of FSW (alongside other welding processes such as laser welding) to replace many of the rivets in current structures. The first aircraft to make extensive use of FSW in its airframe, the Eclipse 500 business jet, has recently completed certification and is now in production.^[92] In this aircraft, over 7300 fasteners (approximately 60% of the total) are replaced by 263 friction stir welds.

Rail

High speed aluminium railcars such as the Japanese Shinkansen are normally built from complex double skin extrusions in 6xxx alloys.^[93] Since the welds which join these are long (up to 25m) and straight, FSW is an ideal process, and the very low distortion is cited as a major benefit^[94,95] (see also Fig.2a).

Automotive

There are few long straight welds in road vehicles, and so adoption of FSW has primarily been for components such as suspension parts, wheels, seat components, crash boxes, etc. where several leading companies are already using the process in production. The needs of the automotive sector have driven the development of robotic FSW, to cope with the complex shapes and high volume/low cost culture of this market. Significant interest is now being shown in friction stir spot welding, where the linear translation of the tool is either very small or zero. Friction stir spot welding is rapidly gaining acceptance as an efficient method of joining aluminium sheet, and is already in production, for example on the Mazda Rx-8 sports car, where it is used on the aluminium bonnet and rear doors.^[96] Friction stir welding is also being developed for lightweight armoured vehicles, where the ability of the process to weld material of around 25-40mm thickness in one pass is being exploited.^[97,98]

Process variants

In the past five years, two variants have emerged as significant technologies in their own right, namely friction stir processing^[74,99-102] and friction stir spot welding.^[103-105] Mishra et al.,^[101] Mahoney et al.^[106] and Charit et al.^[107] have demonstrated the capability of the former to produce fine scale microstructures in superplastic 7xxx alloys, greater than 5mm thick. Similar results have also been demonstrated in an Al-Li-Cu alloy^[108] and in commercial purity aluminium alloy 1050.^[109] The technique has also been demonstrated as effective for homogenising powder metallurgy processed alloys.^[102,110] The ability to process aluminium alloys in this manner raises a range of possible applications for friction stir processing. These include tailoring microstructures for subsequent deep drawing and superplastic forming operations, and the ability to refine locally the microstructure of castings (for example, around stress concentrations, where a superior wrought microstructure would be preferable). Friction stir processing is not considered further in the present review.

Friction stir welding process

Flow mechanisms and tool design

The metal flow and heat generation in the softened material around the tool are fundamental to the friction stir process. Material deformation generates and redistributes heat, producing the temperature field in the weld. But since the material flow stress is temperature and strain rate sensitive, the distribution of heat is itself governed by the deformation and temperature fields. In fact their control lies at the core of almost all aspects of FSW, for example, the optimisation of process speeds and machine loading, the avoidance of macroscopic defects, the evolution of the microstructure, and the resulting weld properties.

As noted above, almost all the material in the weld is extruded between the rotating pin on the retreating side and the surrounding material which is too cold and too lightly stressed to deform (see Fig.1). In its simplest form, this essential flow mechanism can be illustrated by two-dimensional simulations depicting streamlines round a rotating tool placed in a steady flow of material. Figure 3a shows streamlines past a cylindrical tool, redicted by computational fluid dynamics (CFD).^[111] A longitudinal weld seam is formed behind the advancing edge of the tool where the two flows come together. Further modelling studies have investigated how this two-dimensional flow is perturbed by:

1. the addition of tool features such as flats and flutes
2. changes in the contact conditions between tool and workpiece, from sticking friction to slipping at a lower interfacial shear stress.

Predicted streamlines round a fluted tool are shown in Fig.3b and c.^[112] Complete sticking generates a dead metal zone round the tool, whereas the flow interacts closely with the tool features when slipping takes place. Another characteristic of the process - a line initially perpendicular to the welding direction is swept into a backwards 'bulge' in the wake of the tool - can also be seen in Fig.3b and c. Marker experiments^[113] have confirmed this behaviour (Fig.3d). One way of quantifying the mixing effect of the tool is the ratio of the swept volume to the pin volume.^[114] For 25mm thick plates this has found to be 1⋅1 : 1 for a cylindrical pin, 1⋅8 : 1 for the Whorl and 2⋅6 : 1 for the MX-Triflute pin (see Fig.2c), each having similar root diameters and lengths^[114] with the Triflute giving the more parallel sided weld zone. Further refinements include the Trivex tool^[115] which was designed to reduce the down and traverse forces required and the Triflat tool for thicker section materials (Fig.2d).

Fig.3. a) typical generic flow path of plate material round clockwise rotating pin in FSW, taken from two-dimensional CFD model with cylindrical tool moving from left to right (after Seidel and Reynolds),^[111]

b, c) effect of interfacial boundary conditions (b stick; c slip) on predicted flow from two-dimensional CFD model with profiled tool (after Colegrove and Shercliff):^[112] change in thickness of streamlines indicates final location of points initially forming straight line transverse to weld line (analagous to Cu foil in d) and

d) metallographic marker experiment using transverse copper foil, illustrating flow induced by pin (after Reynolds) ^[113]

Experience has shown that it is advantageous to develop a vertical component to the flow of material, and most tools therefore contain threads, helical flutes, or similar features to force material adjacent to the pin to flow away from the shoulder. Further variants have emerged as the process matures, such as tools in which the pin and shoulder rotate independently (including non-rotating shoulders),^[33] retractable pin tooling,^[94] as well as a bobbin tool with a shoulder on both ends of a pin of length equal to the plate thickness.^[116] Tool design has been reviewed in detail by Fuller^[117] and Dubourg and Dacheux.^[118]

The capture of three-dimensional flow greatly complicates the modelling challenge.^[78,119] This is not simply because of the geometric complexity, but also because of the inherent sensitivity in the flow response to the interfacial conditions and the temperature and strain rate sensitivity of the material flow stress. A further feature of the flow that is not captured by current models is the formation of a stable void immediately behind the tool (see the section on 'Formation of voids'). This has been identified by 'stop-action' experiments, in which the traverse is abruptly arrested, with tool extraction being synchronised with rotation to preserve the material in contact with tool features. An early example of this technique is shown in Fig.4.^[120] The threads on the leading edge were full and left intact, while on the trailing edge the upper threads are incomplete. The conditions that lead to formation of this void, and whether it has any relevance in producing a sound void free joint, remain open research questions.

Fig.4. Longitudinal vertical section in travel direction after synchronised pin retraction, showing trailing void on left hand side (after Colligan) ^[120]

Heat generation and process operating regimes

Friction stir welding differs from competing processes such as arc and laser welding, since these use an external heat source of specified power, whereas in FSW the joining process itself generates the heat. The heat input is therefore a complex function of the process variables (traverse and rotation speeds, and down force), the alloy being welded, and the tool design. The effect of tool rotation and traverse speed on the heat input per second and per mm are shown in Fig.5a and b respectively.^[53] Analytical estimates of heat input have assumed sliding Coulomb friction at the tool/workpiece interface with a constant coefficient of friction, or sticking friction using an estimate of the limiting shear yield stress, or have inferred contact conditions and/or heat input from measurements of machine torque.^{[77,78,121-127]} Thermocouple measurements coupled to heat flow analysis also provide a means to infer net power inputs. However, Peel et al.^[53] found no simple correlation between the temperature and the input power or heat. Although the heat input is commonly considered in fusion welding, it is a poor indicator of the temperature of the material surrounding an FSW tool, at least for the joining of thin plates of aluminium. It is likely that when the traverse speed is reduced, much of the additional heat is conducted into the backing plate, as evidenced by the observed correlation between the heat input and the backing plate temperature^[53] as well as through the tool. The down force in FSW provides intimate thermal contact between workpiece and backing plate, but this contact evolves with position during the welding process, requiring complex calibration.^[78,127-129] The growing recognition of the importance and complexity of heat input has lead to the routine instrumentation of welding equipment, with spindle rotation speed and torque measurements now providing detailed power data.

Fig.5. Rate of heat input a per millimetre of weld line and b per second for like to like and AA5083/AA6082 welds ^[53]

Heat is produced primarily by viscous dissipation in the workpiece material close to the tool, driven by high shear stresses at the tool/workpiece interface. The temperature and normal contact stresses vary widely over the tool, so it is unlikely that a single contact condition will be valid. Material at the interface may stick or it may slip, or there may be a combination of the two. As discussed above, local melting may occur as peak temperatures reach the solidus temperature. There may be then be oscillating stick slip behaviour, as local melting rapidly reduces the shear stress, leading to a steep drop in local heat input and temperature, and self-stabilising behaviour. Process modelling using CFD has been used to explore the sensitivities of the heat generation, tool forces and size of deformation zone as a function of tool design and process conditions.^[78] Recently the heat input has been calculated directly from the hot deformation constitutive response of the alloy, using a fully coupled deformation and heat flow model.^{[14, 130]} This approach is currently limited by the quality of the constitutive data, particularly near the solidus, but has the potential to reveal the deformation regime that corresponds to the production of sound welds with least machine load for any particular alloy. For example, Fig.6 shows the predicted variation of power generation, peak temperature and traverse force with rotation speed. Note that the power and peak temperature saturate as the interface temperature and heat generation are limited by rapid material softening, while the minimum force required is predicted to occur at an intermediate rotation speed. This result correlates with very early Russian work on rotary friction welding by Vill,^[131] which shows that the time to complete a rotary friction weld goes through a minimum as the rotation speed is increased. The effect has been confirmed experimentally by TWI.^[132].

Fig.6. Schematic illustration of variations, predicted by CFD, of heat generation, peak temperatures and traversing force with rotation speed for FSW of 2024 aluminium alloy (after Shercliff and Colegrove) ^[78]

For a given alloy and plate thickness, with a particular tool, the operator's remaining process variables are down force, tool tilt angle, tool plunge, rotation speed and traverse speed. Down force is only a preset variable when in force control, and plunge depth is only a preset in position control. Empirical trials invariably explore a matrix of these variables, thereby defining the process window that produces sound welds without tool breakage or macroscopic defects. Figure 7 shows the relationship between the welding parameters and the FSW process window for an aluminium die casting alloy.^[133] With increasing tool down force the process window is enlarged mainly towards lower rotation speeds and higher welding (traverse) speeds. This pattern is broadly typical of aluminium alloys generally. The process operating window is commonly described as being limited by 'hot welds' and 'cold welds' - the former associated with high rotation and low traverse speeds, leading to excessive flash production, the latter with low rotation and high traverse speeds, leading to tool breakages. The nature of the defects associated with unsound welds is discussed further below.

Fig.7. Range of optimum FSW conditions for various tool plunge down forces for 4 mm thick ADC12 Al-Si casting alloy welded using 15 mm shoulder, 5 mm diameter, 3⋅9 mm long threaded pin:^[133] flash style flaws were associated with excessive heat input, whereas tunnel style voids (Fig.11) were associated with insufficient heat input and abnormal stirring

Macroscopic weld features

Definition of macroscopic weld zones

The first attempt at classifying FSW microstructures was made by Threadgill.^[134] This work was focused solely on aluminium alloys, and was limited to features distinguishable by light microscopy. However, work on other metallic materials has demonstrated that the behaviour of aluminium alloys is not typical of most metals and alloys, and this initial classification was inadequate. Consequently, a revised set of terms was suggested^[135] and then subsequently revised^[4] and adopted in the American Welding Society Standard D17⋅3M.^[136] These microstructural terms are illustrated in Fig.8, and are defined below along with alternative terms commonly found in the literature:

1. unaffected material or parent metal: material remote from the weld, which has not deformed and which, although it may have experienced a thermal cycle from the weld, is not affected by heat in terms of detectable changes in microstructure or properties
2. heat affected zone (HAZ): the region close enough to the weld for the weld thermal cycle to have modified the microstructure and/or properties, but no apparent plastic deformation is detected by light microscopy although it is recognised that some plastic deformation will have occurred, as is typically the case in any weld HAZ (some researchers have preferred the term 'thermally affected zone'; however, by analogy with other welding processes, there is little justification for a distinct terminology for friction stir welds where the term HAZ is well understood)
3. thermomechanically affected zone (TMAZ): in this region, the material has been plastically deformed by the FSW tool, and heat from the processing has also affected the material. In the case of aluminium, it is possible to generate considerable plastic strain without recrystallisation in this region, and there is generally a distinct boundary, at least at a macroscopic level, between the recrystallised and deformed zones of the TMAZ. (The recrystallised region is often called the 'nugget', which is a descriptive term, though not very scientific. Terms such as 'dynamically recrystallised region' have been suggested^[136] and used extensively in the literature, but the term nugget remains widely used and understood).

When presenting micrographs it is also conventional to show the advancing side of welds on the right and this convention is used here except where specific mention is made in the figure caption.

Fig.8. Microstructural zone classification in a friction stir weld in Al 2024 alloy (A: parent material, unaffected by process; B: HAZ, thermally affected but with no visible plastic deformation; C: TMAZ, affected by heat and plastic deformation)

Mixing across dissimilar welds

Friction stir welding has been used with notable success to join dissimilar aluminium alloys in a number of configurations (for references, see the introductory section). It would be reasonable to assume that a process with so much shear strain would result in very effective mixing of the alloys, but experience has shown this is seldom the case. The 'handedness' of the weld is an important factor, i.e. which alloy is placed on the advancing side in a dissimilar combination.^[53] Both the heat generation and appearance of the weld cross-section change significantly when the weld handedness is reversed, for the same rotation and traverse speed. Figure 9 shows macrographs of the stir zone in the dissimilar 6082/5083 welds produced using a range of rotation speeds, with AA6082 on the retreating and advancing side respectively. Changing the traverse and rotation speeds can have a significant effect on the flow of material within the stir zone. Generally speaking, the extent of mixing and interface disruption increases as the rotation speed is increased, or the traverse rate is decreased.^[53] Unsurprisingly the rotation speed was found to have a significantly greater impact than the traverse speed. For a given combination of weld parameters, the welds produced with AA6082 on the advancing side (Fig.9b) exhibited a significantly lower level of mixing in the stir zone than those with the materials reversed. This is in contrast to the results of Larsson et al.,^[49] but corroborates those obtained by Tanaka and Kunagi.^[50] Further examples are shown in Fig.10, which shows examples of Al-Mg and Al-Si castings welded to 5xxx and 6xxx series plates respectively. The alternating bands of material originating from the casting and the plate material are clearly seen. It has been suggested that the distance in the welding direction between dissimilar bands corresponds with the pitch distance (the travel speed divided by the tool rotations per second). ^[137]

Fig.9. Macrographs showing stir zone/TMAZ of 6082/5083 dissimilar welds with a 6082 on retreating side (left side lighter);

b) 6082 on advancing side (right side lighter) for welds made at traverse speed of 200 mm min21 and various rotation speeds (dotted lines on 560 rev min21 welds show approximate size and position of 6 mm dia. pin)53

a) AA6082 on retreating side (left side lighter);

b) 6082 on advancing side (right side lighter)

Fig.10. Macrosections of friction stir welds

a) Al-Mg-Fe casting (CC601-T6) to 5484-T34 alloy;

b) LM6 Al-Si-Mg casting to 60682-T651 alloy

Flaws in friction stir welds

Flaws arise in most materials joining processes. For example, when arc welding aluminium alloys, weld metal porosity^[138] and, depending on the particular alloy, weld metal solidification cracking and HAZ liquation cracking^[139] are among the most common flaw types. The occurrence of such problems has contributed to the widely held view that some aluminium alloys, in particular some of the high strength 2xxx and 7xxx series alloys, are difficult, or indeed impossible, to fusion weld successfully. Being a solid state joining process, FSW obviates the problems of porosity and hot cracking. In this respect it is worthwhile to make a distinction between flaws and defects, although the two terms are often used interchangeably within the literature. The usual distinction is that a flaw or imperfection is a feature that one would prefer not to be in the weld, but it may or may not compromise the integrity of the weld. If, after evaluation, the flaw is deemed unacceptable, then it becomes a defect. If it does not compromise the integrity, then it is a tolerable flaw. Flaws or discontinuities should be characterised as defects only when specific acceptance criteria, related to the engineering application, are exceeded, and the presence of the flaw compromises the integrity of the structure. Table 2 summarises the characteristic flaw types in butt and lap welds in friction stir welds and their principal causes. In fact the most common flaw types are caused by use of under optimised parameters or a lack of process control. Since understanding of the causes of these flaws/defects is good, it is usually possible to rectify these problems by changes to parameters, tool designs or operating practice.

Formation of voids

Figure 11a shows a typical void on the advancing side of a weld. Similar features are sometimes observed near the base of the pin (Fig.11b). The formation of a continuous tunnel void in this location was a common observation in early FSW, but was eliminated by the use of tool tilt or the redesign of pin features. Figure 11c shows a near surface defect left by the trailing edge of the shoulder.^[140] Since voids are not in general surface breaking, they cannot usually be detected by visual or surface inspection but are relatively easy to pick up by NDE (see below). A number of factors have been identified as contributing to void formation, including inadequate welding pressure,^[140] high travel speed and slow tool rotation speed^[141,142] as well as inadequate control of the joint gap.^[140,143] If the welding pressure is inadequate, the weld will receive insufficient forging action from the tool shoulder to achieve full consolidation.^[140,144] When welding at high travel speeds, and also slow tool rotation speeds, the material receives less work per unit of weld length, i.e. fewer tool rotations per millimetre. Under such conditions, the plasticised material may not reach a sufficiently high temperature.^[142,145] One view is that aluminium alloys can withstand only a certain shear strain rate, which is dependent on temperature. When subjected to lower shear strain rates, they will flow and recover, but at higher strain rates they cannot flow/recover fast enough to keep flowing, and they will break up, forming voids, etc. Colegrove and Shercliff^[130] suggested that the void defect shown in Fig.11b formed near the base of a Trivex pin was due to inadequate forging and consolidation due to the low temperatures there. Kumar et al.^[146] found that a rounded tool was less likely to produce void defects.

Table 2 Summary of characteristic flaw types encountered in friction stir welds

Fig.11. Characteristic void flaws in friction stir welds:

a) volumetric flaw in 2014;^[140]

b) tunnel (wormhole) defect at base of Trivex tool when welding 7449 at 120 rev min^-1/60 mm min^-1; ^[130]

c) surface defect located under shoulder in 2014A (Ref.^[140])

The presence of voids on the advancing side of the weld (Fig.11a) has been investigated by mathematical modelling of metal flow during the welding process.^{[9, 119, 130]} These models predict that transitional volumes of material will form, between regions of rotational motion of material and regions in which material is extruded past the rotating tool pin. Bendzsak et al.^[9] described the motion in the transition region as 'chaotic'. The models predict flow singularities (stagnation) in this region of the advancing side of the weld, which have been considered to be the source of tunnel defects (Fig.11b).^[130] These defects form under inappropriate welding parameters, such as high welding speed or low pressures. Further increasing the welding speed or decreasing the welding pressure would cause extension of the tunnel defect up to the weld surface, ultimately forming a groove type defect.^[147] It is noticeable that the tunnel defect is often observed near the bottom of the weld at the advancing side. From these observations, FSW could be considered as a process where a cavity formed behind the welding tool is later filled by plasticised material flowing from the front of the pin to its rear (see also Fig.4). If the cavity is not filled, then a cavity will remain in the weld. This explanation is supported by the study of Zhao et al.^[148] who found that under identical welding conditions tapered and cylindrical pins with threads did not show wormhole defects while those without did, because the thread helps to transport material around the tool back to the advancing side, leaving a void there. Crawford et al.^[149] related the lack of adequate flow and wormhole formation to low weld pitch, i.e. insufficient rotational speed relative to the weld travel speed.

Prediction of void formation is a particularly difficult modelling problem, due in part to the limitations of the numerical methods used for flow modelling. Computational fluid dynamics solvers treat the deforming metal as a hot, viscous fluid, neglecting elasticity. As noted above, CFD gives a good representation of flow patterns and the internal generation of heat, but cannot readily describe a free metal surface - the deforming metal fills all the space between the tool and the backing plate. Void formation necessarily implies a free surface, and is also strongly influenced by the hydrostatic pressure and the mechanism of cavitation. Computational fluid dynamics models indicate that a region of hydrostatic tension forms on the tool wake on the advancing side, but the neglect of elastic stresses means that the predicted pressures are unreliable quantitatively, and there is in any case no criterion available for the formation of a stable void in hot deforming metals.

Finite element methods are generally better suited to elastic-plastic analysis and therefore to predicting residual stress (see below), but not to material flow because the strains are very high, due to the computational demands of continuous, fine scale remeshing.^[150] Some success has been achieved in modelling FSW using the arbitrary Lagrangian-Eulerian formulation.^[151,152] However, computational run times mean that only very short weld traverses can be simulated, and the models have great difficulty in capturing the complex weld closure behind the tool, due to the sensitivity to the local stress state of separation between workpiece and the tool.

Joint line features

Joint line remnants are features that extend from the weld root, at the location of the original parting line between the butting plates to be welded, through the weld (Fig.12a). Joint line remnants distributed through the bulk of the weld have been reported by several workers in many alloys.^{[140,153-157]} In some instances, such features can be little more than a string of inadequately dispersed oxide particles, which originated from the surfaces of the butting plates. A recent study by Sato et al.^[158] using transmission electron microscopy has confirmed that these strings are indeed Al₂O₃ particles. These are a consequence of the presence of the natural aluminium oxide on the surfaces of the butting plates.^[154,155] The original joint line is clearly discernable in Fig.12a, while at higher magnification, the clusters of oxide along the prior interface are visible (Fig.12b). Such features are affected by the welding speed; increasing the welding speed reduces the disruption of the oxide per unit advance of the tool. Tool shoulder size and tool geometry^{[140,154,155]} may also have an effect. In many cases such bands of oxide can be innocuous provided good mixing occurs.^[158,159] Reynolds^[113] suggested that the curving line arises from remnants of the oxide layer, because the final position of the initial butt surface after material flow of FSW exhibits a curving line on the cross-section. This has been verified using marker material experiments in which a thin copper layer is trapped in the joint line, and its final position identified using X-ray tomography (Fig.3d).^[160] Figure 13 shows a tomograph of a stop action weld, illustrating the breakup and redistribution of the joint line. In cross-section (lower image), it can be seen that the copper is predominantly distributed in a characteristic curve bordering the weld nugget on the retreating side.

Fig.12. a) oxide defect in 5083 alloy which can in severe cases lead to 'kissing bond' at the base,^[161]

b) magnified view showing increased oxide inclusion level in 2014 along prior joint line^[140]

c) schematic of joint line remnant ^[161]

Fig.13. X-ray tomographic and corresponding metallographic interpretation of FSW flow mechanism (after Dickerson et al.):^[160] three-dimensional X-ray tomography image (top) showing break-up of Cu foil placed on joint line; bottom: virtual cross-section revealed by tomography

From the above discussion, it is important that the oxide interface between the butting plates is adequately disrupted in order to form a bond, and in the vast majority of cases this is exactly what happens (Fig.12c). Indeed the original joint line is normally increased in length some 3-5 times, and the oxide is broken up and scattered, so that most of the original joint line is a metal to metal bond, and therefore strong. In some cases, the interface is only partially disrupted and remains identifiable in a cross-section. It is unusual for such joint line remnants to be a problem.

Naturally, the correct depth of penetration of the tool pin is essential to ensure that mixing occurs over the full plate thickness. Where a shortened pin is used, where the tool plunge depth is incorrectly set, or where there is poor alignment of the tool relative to the joint line, a root flaw can be produced, indicative of a complete lack of bonding (Fig.14). In such cases the bond quality improves towards the nugget. As a result there can be a transition region where a bond exists, but is weak. This is generally known as a 'kissing bond', a rather unscientific name which has become widely adopted. It is not possible to identify a kissing bond from a microstructural examination. Zhou et al.^[161] have found that kissing bonds can show 20-40 times shorter fatigue lives for Al 5083 welds and 30-80 times shorter lives for Al 2024 welds. The fatigue fracture surfaces observed by SEM were consistent with a kissing bond in that the fractured specimens failed from the root tip of the oxide array. An example of failure from such a kissing bond under tensile loading is depicted in Fig.15. It should be emphasised that kissing bonds are rare and can be eliminated by appropriate tool/joining parameters.

Fig.14. Characteristic root flaw in FSW 2014A caused by using too short a pin in 5083 alloy

Fig.15. Cross-section of 5454 FSW specimen with crack after tensile test along kissing bond ^[156]

Other flaw types

Flash style features can occur on the top surface: excessive flash is not necessarily a bad thing; in some products high flash levels are deliberately introduced to ensure adequate penetration, especially in welds where the fit-up is poor. Flash is normally caused primarily by large plunge depths.

An unusual 'chevron' surface feature has been observed on the root surface of welds in 5xxx series alloys (Fig.16). Pryzdatek^[162] noted that the features were coincident with small root line imperfections in welds in alloy 5083 and concluded that they were the result of the expulsion of material from the joint during welding lending evidence to the possibility that some liquation may occur during welding. Leonard^[163] observed similar features in both 5083 and 5251 alloys. To date the features have not been reported to affect either the mechanical or corrosion behaviour of the joints and it is possible that the features will remain a curiosity rather than a point of concern. Pryzdatek^[162] has noted, in any case, that light dressing will remove them.

Fig.16. Chevron markings on underside of friction stir weld in 5083 alloy (scale in mm)

Local melting during FSW has been a topic of continual debate. However, conclusive evidence of local melting in aluminium alloys has been rarely reported. Figure 17 shows a region of the TMAZ beneath the tool shoulder of a weld in alloy 7050-T7451, produced at TWI,^[164] providing clear evidence that liquation can indeed occur. While this demonstrates that liquation is possible, the paucity of reported cases of liquation cracking in the literature suggests that the occurrence is less widespread than in aluminium arc welds. The apparent absence of extensive liquation during welding may be explained as a consequence of the extensive mechanical deformation experienced during welding. Any liquated area is likely to be difficult to identify in a metallographic section of the completed weld. It is also worthy of note that liquation has been found in the TMAZ region of friction stir welds in magnesium alloy ZK60 by Johnson.^[165] The formation of grain boundary films was observed, but these could be prevented by using a lower heat input welding procedure.

Fig.17. Evidence of partial melting on retreating side of nugget in FSW of 6 mm thick AA7050-T7451 alloy ^[164]

Sato et al.^[64] have described liquation when welding aluminium alloy 1050 to a magnesium alloy. In this case, the maximum temperature reached can exceed the melting point of the Al12Mg17 intermetallic, with clear evidence of a cast structure being visible. However, this is a special case resulting from the welding of two very different materials which form a low melting point eutectic.

Lap weld defects

Friction stir lap welds may contain some of the flaw types encountered in butt welds, in particular voids and oxide joint line remnants, the latter often appearing in a horizontal orientation through the joint. Owing to the joint geometry, root flaws are not encountered. However, lap welds do exhibit features at the edges of the bonded region which affect joint properties. Any friction stir lap weld, like a riveted lap joint, contains notches at the faying surfaces of the sheets. Cederqvist and Reynolds^[166] have shown that such features in friction stir welds can have a marked effect on joint performance. The notch feature can deviate towards either the top or bottom surface of the weld, depending on welding conditions, reducing the effective sheet thickness. This feature is sometimes described as 'hooking', involving a significant rotation of the unwelded interface, sometimes by 90°C. It is found on the advancing side of the weld. It is also not unusual to find upper plate thinning on the retreating side of the joint, and this can significantly compromise the load bearing capacity of the joint. Typical examples are shown in Fig.18. Thomas et al.^[167] have shown that welding tool design can have a strong influence on the occurrence of such features and consequently joint properties, in particular fatigue. CFD modelling has also been used to show how the tool profile modifies the extent of hooking in lap welds, enabling a degree of prior computer based optimisation of tool profiles.^[168]

Fig.18. Lap weld defects showing hooking on advancing side and plate thinning on retreating side in lap welds between 7075 (upper) and 2024 (lower) alloys

Non-destructive monitoring of defects

Currently, only limited standards exist for evaluating the quality of friction stir welds^[136,162] and further work is required to assess the significance of flaws and their means of detection. It should be noted however that all welds, whatever process is used, are likely to contain some imperfections, but most do not seriously compromise performance. Indeed the overwhelming majority of friction stir welds do not undergo nondestructive inspection, partly because of cost and partly because the incidence of significant flaws is small, and the risk of a catastrophe minute, as most structures have considerable redundancy. As a result, hundreds of kilometres of friction stir welds have been safely employed commercially in most cases, with no more than visual inspection. In areas where a flaw is absolutely unacceptable (e.g. rocket fuel tanks) very detailed NDE is used, although defects are extremely rare. In other cases, a statistical process control method is used. If the QA system is tight, and the parameters fall within certain narrow ranges, then the risk of a defect in a fully mechanised process is extremely small.

Traditional NDE techniques do not work well with FSW because the techniques were developed for fusion welds, where different flaws/defects are found, e.g. porosity, solidification cracks, hydrogen cracks, reheat cracks, lack of fusion, slag entrapment etc., none of which is found in FSWs. The NDE industry is responding by developing/adapting techniques to cope with this challenge. Using techniques such as phased array ultrasonics and meandering winding magnetometers, voids can now be readily detected. A good example of the capability of the ultrasonic method as applied to FSWs is given in Fig.19.

Fig.19 a) Flaws in cross-section from 6061 FSW tailor welded blank specimen (cylindrical threaded pin, welding speed 1200 mm min^-1, spindle speed 1500 rev min^-1, shoulder penetration 0 mm) and

b) corresponding synthetic aperture focusing technique ultrasonic image ^[169]

Current NDE techniques are not totally reliable for detecting root flaws. The only definitive method is a destructive bend test with the root in tension.^[136] Efforts are being applied to both advanced ultrasonic inspection techniques (for example phased array techniques), and sophisticated eddy current techniques (meandering winding magnetometers, pulsed eddy currents).^[169-176]

'Kissing bonds' (see the section on 'Joint line features'), however, are particularly insidious, as they are difficult to detect using non-destructive techniques which rely on an interruption in the microstructure. While the oxide stringers are not directly sensed by ultrasonic attenuation, the associated reduction in grain size can be detected.^[176] By comparing the mean noise level inside the root to that of the weld nugget, the operator can estimate the pin depth and therefore the likelihood of having a kissing bond.

In summary, the FSW process is now mature. Various tools have been designed to meet differing geometrical demands. In most cases relatively wide processing windows have been established for the production of large quantities of defect free welds using automated commercial welding systems.

Microstructural features and hardness profiles

Introduction

In any welding process, the properties and performance of the weld are dictated by the microstructure, which in turn is determined by the thermal cycle of the welding process, which can normally be varied by changing the welding parameters. Therefore welding parameters must be selected that give the best possible microstructure and that allow welds to be made free from defects and other undesirable features. With most materials, it is well understood that welding has some adverse effects on microstructure and properties, and thus the 'optimised' weld parameters are often a compromise between making sound welds at economical production rates and producing acceptable, rather than ideal, microstructures and properties.

Friction stir welds in aluminium alloys contain a wide variety of microstructures, which is hardly surprising when the extreme range of strains, strain rates and thermal cycles to which different regions of the weld are exposed is considered. The microstructural variations were first characterised by Threadgill^[134] (see Fig.8). In the HAZ, remote from the centre of the weld, there is no obvious change to the grain structure (Fig.20c), and the HAZ is detected only by a change in hardness and generally by a change in etching response. In precipitation hardened alloys it is widely accepted that some coarsening of precipitates is occurring, and possible dissolution at higher temperatures. In work hardened alloys, dislocation networks may recover, and this may cause some low angle cell boundaries to form. Furthermore as the weld centre is approached, clear evidence of plastic deformation can be seen in the grain structure. In the outer part of the TMAZ, the original grains remain identifiable in the deformed structure, with the formation of subgrain structures and significant associated rotation of the parent grains as evidenced by the pole figure in Fig.20b. Closer still to the weld line, the strains, temperatures and time at elevated temperature all increase, allowing the formation of the recrystallised nugget with a fine equiaxed structure (Fig.20a). The microstructural characteristics will first be discussed for the nugget region, in which deformation dominates. Evolution of microstructure in the heat affected zone is thermally controlled, and this will be discussed separately for non-heat-treatable and heat-treatable alloys. The microstructure of welds made between dissimilar alloys is also discussed briefly.

Fig.20. Microstructure of 2199 alloy FSW displayed using inverse pole figure map obtained by EBSD, showing refinement of microstructure in nugget

a) nugget;

b) nugget/TMAZ boundary;

c) HAZ region^[177] (scale bar corresponds to 50mm in each case)

Deformation microstructure in weld nugget

'Onion ring' structure

A common observation from the nugget region in FSW is the appearance of a series of circular or elliptical features in etched metallographic sections (see, for example, Fig.8), often termed 'onion rings' (as the sections reveal a slice through a set of nested layers of roughly hemispherical shape, like an onion). The significance of this structure in the weld nugget remains an occasional topic of interest in the literature. Mahoney et al.^[6] and Leonard^[178] have shown for alloys 7075 and 2014A that the ring patterns are an etching response to variations in grain size between the rings. Other characteristics of the rings include texture effects^[179,180] and variations in dislocation density.^[181] The nugget may also contain fractured constituent particles^[178,182] and the structure has been attributed to a variation in their distribution.^[178,183] This is turn may be a consequence of the banded distribution of the constituent particles present in the base metal, a characteristic that is strongly alloy dependent.^[184] These factors primarily relate to the strength of contrast in microstructure observed in the weld nugget, but do not offer a complete explanation of the mechanism of formation, which has not yet been formulated. There seems to be strong argument that there is a purely kinematic basis for the formation of each ring, associated with one rotation of the tool (or the rotation between positions of tool symmetry, i.e. three per revolution for a Triflute tool). Cyclic fluctuations in the amount of material extruded past the tool and being deposited are to be expected with profiled tools.^[185] It has therefore been postulated that ring formation may be a function of the tool geometry, tool rotation and forward travel speeds.^[134] Computational fluid dynamics modelling^[112] and marker experiments^{[120, 186]} have also made a modest contribution to the discussion to date. The practical significance of the phenomenon remains rather limited as the mechanical properties of the nugget are generally good, and the fracture paths in mechanical tests are seldom associated with the onion rings.

Recovery versus recrystallisation

A feature of the microstructure of friction stir welds in aluminium alloys is the development of a fine grain structure in the centre of the nugget region, as shown in Fig.20a. Typically, equiaxed grain sizes of the order of a few micrometres have been measured in the nugget region. The precise nature of this fine grained region has been the subject of much research and debate in the literature. TEM studies by several workers, examining a range of different alloys, have shown that the fine grain structure in this region in the light microscope comprises fine grains possessing predominantly high angle grain boundaries and a low dislocation density.^{[6, 187,188]} On the basis of these observations, it has been concluded that the nugget consists of dynamically recrystallised grains, and not subgrains.^{[6,8, 187,189]} A similar observation has also been made in a dissimilar weld between copper and alloy 6061,^[71] in which fine recrystallised grains rich in both aluminium and copper were observed. One study of alloy 6061-T651, has contrasted the grain structure of the nugget region with that of the remainder of the TMAZ.^[189] Tilting studies showed that the nugget comprised dynamically recrystallised grains, whereas the remainder of the TMAZ comprised deformed subgrains, separated by low angle grain boundaries. Rhodes et al.^[190] proposed that the final equiaxed nugget grains in 7050 are formed by grain growth from much finer grains nucleated by the dynamic recrystallisation process, thus accounting for the low dislocation density. It is also likely that, before recrystallisation, extensive recovery occurred, as there will be significant plastic flow in the material about to be welded. However, other studies^[191] of 7050-T7451 have shown a high dislocation density in grains in the nugget, and a study by Sato et al.^[192] of 6063-T5 showed that whilst most grains exhibited a low dislocation density, some grains exhibited a much higher density. These variations in dislocation density may be associated with the welding process conditions, in particular the forward tool movement per revolution, but this hypothesis has not been tested. High values of forward tool motion per revolution produce harder microstructures, but generally similar grain size. The influence of grain size and dislocation density on strength is difficult to isolate in the heat-treatable alloys, due to the simultaneous effects of changes in precipitation hardening, requiring detailed TEM studies. Furthermore the presence of precipitates as a direct influence on the processes of recovery and recrystallisation. Hassan et al.^[193] attempted to simulate the formation of the nugget in AA7010 alloy using high strain rate torsion tests. This study confirmed the effects of a high precipitate density in increasing the resistance to recrystallisation, which only occurred at strains of >20. The process of recrystallisation was also aided by heterogeneous plastic flow.

In a study carried out in the region of the tool pin exit hole in a sample of alloy 7475, it has been argued that the structure of the weld nugget is one of dynamically recovered subgrains.^[194] This was argued on the basis of rapid and massive grain growth observed during annealing experiments at temperature above 500°C.

This postulation that grain growth is due to the presence of a dynamically recovered subgrain structure in the TMAZ has been supported by a study of friction stir welds in alloys 6082-T6 and 7108-T79, using scanning electron microscopy and electron backscattered diffraction.^[195] Using this technique, which can quantify grain orientations on a polished surface, low angle boundaries were reported, indicative of a subgrain structure. Fonda et al.^[196] have also presented evidence from stop action tests which indicates that recovery based mechanisms could be of importance in grain refinement during FSW, although they do not dismiss the possibility of recrystallisation. However, Mishra et al.^[197] have reported abnormal grain growth during heat treatment of friction stir processed alloys 7050 and 2519 in the temperature range about 450-470°C. The authors attributed the phenomenon not to the presence of a subgrain structure, but to a number of possible factors. These included dissolution and growth of precipitates, regions of localised strain differences, regions of non-uniform grain size distribution and the existence of boundaries with different mobility. However, it was recognised that the exact origin of the phenomenon was not clear. Abnormal grain growth is discussed in more detail below.

It has been suggested that the differences in microstructural observations may be resolved by considering a mechanism of continuous dynamic recrystallisation in the TMAZ.^{[182,191,198]} The deformation process associated with welding introduces a large quantity of dislocations, while at the same time grain growth occurs as the temperature rises. Subgrains, which are very small and exhibit low angle boundaries, begin to form by a process of dynamic recovery. Continuous dynamic recrystallisation then occurs as dislocations are continuously introduced to the subgrains by further deformation. The subgrains grow and rotate as they accommodate more dislocations into their boundaries, forming equiaxed recrystallised grains with high angle grain boundaries. Plastic deformation continues with the repeated introduction of dislocations and the process continues until the end of the thermomechanical cycle, at which point partial recovery takes place. The precise mechanism remains unresolved, as recent experiments in which the welding tool was retracted rapidly and the material quenched have shown that very fine recrystallised grains, of the order of 25-100 nm, from Ref.^[190]. These are smaller than the 2-5µm grains observed in the weld nugget under normal welding conditions, suggesting that these arise from a nucleation and growth mechanism. On the basis of such experiments, Prangnell and Heason^[199] suggest that there is no evidence of continuous recrystallisation by grain rotation, but rather bands of fine nugget scale grains are first formed, from closely spaced parallel high angle grain boundaries that develop from finer scale deformation bands. A mixed microstructure develops comprising a matrix of nugget scale grains containing high aspect ratio fibrous grains having more stable orientations. Finally, even these fibrous grain fragments become unstable when they thin to subgrain dimensions and break up to form a full nugget-like microstructure comprised of low aspect ratio ultrafine grains. This mechanism would explain the bands of similarly oriented fine grains in Fig.20b.

Another point which has been debated is whether all of the recrystallisation in the nugget occurs during the deformation process (i.e. dynamic recrystallisation) or whether the process continues after deformation has ceased. The answer is probably more of academic interest than practical significance, but it is reasonable to assume that static recrystallisation (and subsequent grain growth) is more likely to occur in thick section welds, where time at elevated temperature will be longer.

Despite some evidence for subgrain formation, the consensus is that recovery is more important in the highly deformed areas of the TMAZ outside the nugget, where the original grain structure is retained. The fine equiaxed microstructures in the nugget are the result of recrystallisation processes, presumed to be predominantly dynamic.

Effect of post-weld heat treatment on grain structure

As noted above, a feature observed in many alloys is that of massive grain growth in the nugget area during post-weld heat treatment. This not only serves as a guide to the interpretation of the nugget microstructure that leads to this behaviour, but may also have practical consequences for welds subjected to post-weld heat treatment. The phenomenon has been reported in 1xxx,^[200] 2xxx,^[201-203] 6xxx,^{[121, 204,205]} and 7xxx^{[201,206,207]} alloys. However, the mechanism may differ between alloys, as the 1xxx alloys are not precipitation hardened.

Sato et al.^[200] have observed that massive grain growth in 1100-H24 alloy occurs only when the post-weld heat treatment temperature exceeds the maximum temperature experienced during welding, and may be associated with the formation of small grains with high angle boundaries as a result of primary recrystallisation. Typical examples are shown in Fig.21, from work by Litwinski^[202] on 6⋅4 mm 2195-T8A3 alloy. Solution treatment at 510°C (950°F) after welding resulted in massive grain growth in the lower half of the nugget, and also some less spectacular growth just below the upper surface. The problem was solved by modifying the welding procedure to produce a higher welding temperature which reduced the amount of cold work. Attallah and Salem Hassan^[203] showed that the risk of rapid grain growth in 2095 was reduced by high rotation speeds and lower travel speeds, both of which will increase the heat input. These authors observed rapid grain growth both at the upper surface, and in the lower part of the weld.

Work by Hassan et al.^[207] on 7010-T7651 alloy showed that post-weld heat treatment could lead to massive grain growth of generally similar appearance to that shown by Litwinski,^[202] i.e. concentrated in the lower portion of the weld. Hassan et al. have suggested that very fine grain sizes in the as welded nugget are more at risk, as there is insufficient dispersoid available to stabilise the grain boundaries. Grain growth may also be encouraged by the dissolution of these dispersoids during heating. The use of a high heat input welding cycle will give coarser as-welded grains, which should be more stable, although conditions can exist where rapid grain growth can still occur, and it is suggested that this is associated with the formation of planar fronts on the growing grain. It is also suggested that a mean grain size of at least 10 µm would be required to provide grain stability during solution heat treatment. Both Hassan et al.^[207] and Attallah and Salem Hassan^[203] have related the risk of rapid grain growth to theories of cellular microstructures proposed by Humphreys.^[208-210]

Although the precise mechanism may not be fully understood, the above studies indicate strongly that the risk of rapid grain growth during post-weld heat treatment can be reduced by using high heat input welding procedures. These will help to remove cold work and increase the as-welded grain size, and hence the efficiency of a finite number of grain boundary pinning particles. Presumably minimising post-weld heat treatment times and temperatures will also be helpful.

Fig.21. Microstructure in 6 mm thick 2195 alloy a before and b after post-weld heat treatment ^[202]

Grain size control

Certain alloying elements can be added to aluminium alloys to restrict grain growth during high temperature operations, notably scandium and zirconium. There are claims that adding a small quantity of scandium to aluminium alloys will improve fusion weldability (resistance to hot cracking, etc.), with the presence of the thermally stable Al₃Sc precipitate limiting grain growth. Likewise, Al₃Zr precipitates have also been identified as beneficial. There are few published data on FSW of scandium bearing alloys but there is some variance among observations of the effect of Sc on nugget grain size. Gittos and Bridges^[211] (studying two Al-Zn-Cu-Sc alloys, similar to 7010 and 7050), Huneau et al.^[212] (studying an Al-Mg-Sc alloy), and Paglia et al.^[213] (studying cast 7050 containing Sc) found only marginal effects on nugget grain size in the as-welded condition. The Sc addition was ,0⋅12% (Refs.^{[211 and 213]}) or 0⋅26% (Ref.^[212]) and Zr was present to a similar level in the work of Gittos and Bridges^[211] and Huneau et al.^[212] However, unpublished work has shown a beneficial effect of scandium on grain size in the weld. Sato et al.^[214] found that in a binary Al-Zr alloy, the presence of Zr had no significant effect on the nugget microstructures, implying that recrystallisation occurs above the Al3Zr solution temperature. However, the presence of the Zr in solid solution limits dislocation movement and hence recovery mechanisms in the non-recrystallised TMAZ, thus limiting grain growth. Charit and Mishra^[215] have shown very fine grain structures in a friction stir processed Al-Zn-Mg-Sc casting, quoting a mean grain diameter of 0⋅68 µm. This value is significantly less than found in friction stir welds in other alloys, and supports the suggestion of Rhodes et al.^[190] that all welds initially produce such fine grain sizes. This demonstrates that in the absence of the stabilising effect of the Al₃Sc or Al₃Zr precipitate there will be some grain growth. This view is supported by later work from Hsu et al.^[216] who introduced fine dispersions of Al₃Ti into a pure aluminium matrix, achieving mean grain sizes of between 0⋅30 and 1⋅53 µm after friction stir processing. In summary, it seems that the inclusion of grain growth inhibiting elements can be beneficial, but further work is needed to fully understand the mechanisms by which they operate, and to achieve the full potential of these alloy additions.

Texture

A number of studies have examined the development of texture in friction stir welds.^{[153, 196, 217-219]} Studies by Field et al.^[218] indicated that local textures were largely alloy independent, which is supported by the work of others. There are two regimes of texture, one in the area where the process is dominated by the shoulder, and another in the area further down where the shoulder plays no part, as evidenced by Ahmed et al.^[80] for thick section 6082 welds. The texture is very simple in the pin dominated region, as might be expected. Work to date has reported textures in the weld nugget that are consistent with a shear deformation process. Jin et al.^[153] studying welds in 5xxx series alloys, have predicted the deformation characteristics of the weld nugget, concluding that they appeared to be more isotropic than that of the parent plate. This finding has implications for the deformation behaviour of friction stir welds, and for potential applications in, for example, tailor welded blanks. However, the significance of texture development in relation to mechanical properties and sheet forming operations has yet to be rigorously tested.

Microstructure and hardness evolution in nonheat-treatable alloys

Limited microstructural studies have been performed on non-heat-treatable alloys, covering alloys 1100,^[204,220] 5083,^{[180, 221]} 5754, 5251,^[222,223] and 5182.^[153] Welds were made in both the annealed (O) and various cold worked conditions. Macroscopically, welds in these alloys appear similar to welds in heat-treatable alloys, exhibiting a TMAZ and recrystallised nugget.^{[153, 220]} Hardness traverses in work hardened non-heat-treatable alloys (e.g. 5xxx alloys in the H1xx, H2xx or H3xx conditions) normally resemble that shown in Fig.22. As the weld is approached, the heat from the process causes annealing and recovery to take place, leading to a drop in hardness. The minimum hardness is typically in the nugget, where the fine grained, fully recrystallised structure (discussed above) is formed. However, welds made in annealed material (e.g. 5xxx in the O condition) do not exhibit an HAZ. Hardness traces show little or no variation in hardness between the parent metal and weld (Fig.22). Sometimes the weld nugget may be slightly harder than the O condition, due to modest (hot) work hardening and grain refinement. Since the nugget formation eliminates the prior deformation microstructure in cold worked material, the hardness of the nugget region is independent of the original condition, as seen for alloy 5083 in Fig.22.

Fig.22. Hardness traverses across friction stir welds in 5083-O and 5083-H321 alloys showing effect of heat treatment

Since few finely dispersed second phase particles exist to pin grain boundaries, the effects of recovery and recrystallisation are, unsurprisingly, different from those in precipitation hardened alloys. Thus the extremely elongated and deformed grains found in the nonrecrystallised TMAZ in precipitation hardened alloys are not generally seen in precipitate free alloys. The transition from non-recrystallised to recrystallised is far less distinct, implying that recrystallisation is much easier in the absence of precipitates. Experiments by Genevois et al.^[222] have shown that at 350°C, heavily strained 5251 will recrystallise in 15 s, whereas identically treated 2024 required 1800 s to recrystallise.

Microstructural modelling in non-heat-treatable alloy FSW has been limited to predicting the loss of hardness across the weld in initially cold worked tempers.^[52] The minimum hardness corresponds to complete recrystallisation, while the base metal hardness corresponds to no recrystallisation. The problem is therefore to predict the positions between which the volume fraction recrystallised varies from 0 to 100%. The resulting hardness is then estimated using a linear rule of mixtures of the limiting hardness values. The extent of recrystallisation is primarily determined by the peak temperature reached in the weld thermal cycle, together with the duration of the time at temperature. A common approach to modelling microstructural change in a thermal cycle is to replace the cycle with an isothermal hold of duration t_eq at the peak temperature T_p of the cycle. The duration of the hold is defined to be that which provides the same 'kinetic strength', I, as the thermal cycle, defined as

The function t*(T) captures the temperature dependence of the transformation time for isothermal holds (e.g. obtained via saltbath experiments). In practice, it is common to use hardness data from annealing experiments directly, fitting the data semi-empirically as a function of isothermal temperature and hold time. At each position in the weld, T_p is evaluated from a thermal model, the kinetic strength and t_eq found by integrating the equation above, and the values substituted into the hardness function.

Microstructure and hardness evolution in heat-treatable alloys

Heat-treatable aluminium alloys derive much of their strength from the presence of fine precipitates, formed during prior heat treatment (age hardening). The thermal cycles experienced during welding can lead to precipitate coarsening or dissolution, and further precipitation during or after cooling, depending on the peak temperature and duration of the cycle. Given the technological importance of these alloys, FSW has been studied in a wide range of 2xxx, 6xxx and 7xxx series alloys, in naturally aged and artificially aged tempers. The microstructural development of heat-treatable alloys, and the corresponding evolution of hardness (Fig.23), during FSW have been studied extensively.^{[6,8, 48,51, 121, 142, 168, 178,179,182,187-189,191,192,204, 221,222,224-232]}

Fig.23. Schematic plots showing generic hardness responses across friction stir welds

Hardness profiles in heat-treatable alloys

Transverse hardness profiles are a common starting point for interpreting some of the changes that occur during welding. Repeat measurement after a period of natural aging is also useful, indicating that supersaturated solute remains immediately after welding and the final profile may be enhanced. Indeed, since postweld natural aging starts immediately and may continue for many months, reported results are subject to some uncertainty as the interval after welding is frequently not controlled or documented.

Interpretation of hardness data merits a degree of caution for other practical reasons. For example, lower loads may be chosen to enable greater spatial resolution of rapidly varying profiles, but this introduces more scatter. Samples mounted in a hot press may have been further aged by the mounting process. Most published profiles are for the midthickness of a plate, but it is apparent from Fig.8 that variations through thickness will be expected. Only occasional studies map the whole cross-section, due to the effort involved.

Most friction stir welds in heat-treatable alloys, welded in the peak aged or overaged conditions (T6/T7 tempers), exhibit a characteristic hardness profile, as typified in Fig.23. The significant effect of post-weld natural aging for a 7075-T6 weld is illustrated in Fig.24.^[178] In the HAZ, softening is observed, with a rapid drop in hardness as the TMAZ is approached. The greatest recovery in strength is observed in the nugget. Both coarsening and dissolution lead to a drop in hardness, but strength recovery only occurs following dissolution. The hardness profiles are therefore consistent with precipitate coarsening being dominant in the HAZ (lower peak temperatures) and dissolution in the nugget (peak temperatures above the solvus of the initial precipitates), followed by natural aging. The nugget strength may also be augmented by its deformation substructure, as discussed above. For naturally aged tempers (T3/T4) there is the additional possibility that coarsening causes a strength increase in the HAZ, in the region where the peak temperature is comparable with conventional aging temperatures (150-200°C), and the initial GP zones can evolve to a more hardening state. It is also important to recognise that, in all heat-treatable alloys, precipitate evolution is not limited to the bulk of the grains. For example, grain boundaries may exhibit widening of precipitate free zones.^[182,191] This will clearly not be evident in hardness profiles, but may be significant in relation to ductility, toughness, fatigue or corrosion.

Fig.24. Hardness profiles showing natural aging in 6 mm 7075-T6 alloy ^[178]

Faster welding speeds generally lead to colder welds, though there is a more complex relationship between thermal history and speed than in fusion welds, due to the dependence of the heat input itself on the welding speed. Higher speed welds tend to have a narrower HAZ,^[233,234] and the nugget hardness is also often increased due to the reduced deformation temperature and increased strain rate. Figure 25 shows hardness data for 6 mm 7075-T7351 alloy for plates welded at different times during a five year period in which tools and practices evolved significantly: from a traditional threaded pin tool (1995), a traditional tool with accelerated cooling (1997), to an advanced (Triflute) tool allowing a substantial increase in welding speed (2005). A traditional threaded pin tool is now recognised as being of low efficiency. Then the application of a high conductivity toolbed, which extracted heat far more rapidly than a conventional steel bedplate, was applied. More recently, this has been allied to the use of a high efficiency tool (in this case a Triflute tool, see Fig.2). It is clear that progressive reductions in heat have increased the hardness of the nugget, increased the minimum hardness measured at the HAZ/TMAZ boundary, and reduced the overall width of the HAZ.

Since strength recovery is commonly observed in the nugget due to natural aging, it might be expected that post-weld heat treatment would be employed to promote artificial aging. However this is not widely practised on aluminium fabrications, for which a complete resolution heat treatment is not often feasible, or economic. Low temperature heat treatment is practised in some companies to stabilise the microstructure in 2xxx and 7xxx welds. In T3/T4 tempers there is little benefit in elevated temperature post-weld heat treatment. Natural aging potentially restores the weld to the initial strength, if complete dissolution took place in the weld nugget, however the HAZ region will remain softer even after very long periods.

Precipitate evolution in heat-treatable alloys

The detail of precipitation behaviour in friction stir welds is of course more complex than the simple outline above, as revealed by transmission electron microscopy. Early studies[6, 187] on welds in alloys 7075 and 2014 identified that the weld nugget contained overaged precipitates. However, these studies and others[178,191] demonstrate that the hardness of the weld nugget commonly increases during subsequent aging, indicating that the nugget contains sufficient solute capable of sustaining subsequent metastable precipitation and age hardening. Strangwood et al.[187] have postulated, based upon TEM observations of precipitate sizes and volume fractions, that both aging and re-solution of precipitates occur in the nugget during welding of alloys 7075 and 2014. This is further supported by temperature measurements made during welding, which demonstrate that the theoretical dissolution temperatures for hardening precipitates in both 7075 and 2014 were exceeded. Su et al.[182] have argued that the temperatures achieved during the welding thermal cycle for alloy 7050-T651 were sufficiently high to take all of the strengthening precipitates in the nugget into solution. The cooling rate was not sufficiently rapid to prevent reprecipitation, but this was favoured by heterogeneous nucleation at dislocations and constituent particles, resulting in only large precipitates forming. It has been argued that the observed hardness peak in the nugget is due to a number of factors: a decrease in grain size, solid solution strengthening and comminution of constituent particles.[178,187] With regard to comminution, it has been postulated that some dissolution of constituent particles may occur as a result of the thermomechanical processing achieved in the stirred region.[178]

Precipitation studies in the HAZ of alloys 7075 and 2014 have shown full precipitation of the equilibrium phases,[6, 187,191] in common with arc welding.[235] Strangwood et al.[187] and Svensson and Karlsson[188] have shown that precipitation is favoured at matrix/ constituent particle interfaces. Full strength may be recovered only by a full solution anneal;[187] a post-weld aging heat treatment is not applicable.[6, 187] It is worth noting that some aging of the HAZ has been observed in alloys 2014A-T651 and 7075-T651,[178] suggesting that some reversion may take place, but this was within the first few hours after welding, after which no appreciable aging was observed.

Figure 26 shows 7050 welds made in 6⋅35 mm 7050-T7451 alloy. This temper is overaged, designed to give a good compromise between mechanical strength and corrosion performance. More recent TEM studies have systematically investigated the precipitation state in the HAZ, TMAZ and weld nugget, as a function of welding conditions (and also peak temperature of the thermal cycle, inferred from thermal modelling).[236] While TEM studies reveal precipitation behaviour in detail, they have limited quantitative capability, and are restricted practically to a few locations per weld. A powerful technique for obtaining quantitative data covering whole welds is small angle X-ray scattering (SAXS) using synchrotron X-rays. This is an expensive technique and not available routinely, but the studies conducted on FSW offer valuable insight into the process.[230,232,236]

Figure 27 shows representative SAXS maps for precipitate volume fraction and size of η precipitates in 7449 FSW. This technique was used to quantify and explain the differences in final precipitate distributions between T3 and T79 tempers, each welded at high and low speed. The SAXS data were supported by selective TEM observations and provide robust evidence for the interpretation of hardness profiles. In particular, the relative extent of coarsening in the HAZ, and dissolution in the nugget are clearly evident. The higher precipitate fraction in the plate in the T79 condition compared to T3 is also evident.

Since strength recovery is commonly observed in the nugget due to natural aging, it might be expected that post-weld heat treatment would be employed to promote artificial aging. However this is not widely practised on aluminium fabrications, for which a complete resolution heat treatment is not often feasible, or economic. Low temperature heat treatment is practised in some companies to stabilise the microstructure in 2xxx and 7xxx welds. In T3/T4 tempers there is little benefit in elevated temperature post-weld heat treatment. Natural aging potentially restores the weld to the initial strength, if complete dissolution took place in the weld nugget, however the HAZ region will remain softer even after very long periods.

Precipitate evolution in heat-treatable alloys

The detail of precipitation behaviour in friction stir welds is of course more complex than the simple outline above, as revealed by transmission electron microscopy. Early studies[6, 187] on welds in alloys 7075 and 2014 identified that the weld nugget contained overaged precipitates. However, these studies and others[178,191] demonstrate that the hardness of the weld nugget commonly increases during subsequent aging, indicating that the nugget contains sufficient solute capable of sustaining subsequent metastable precipitation and age hardening. Strangwood et al.[187] have postulated, based upon TEM observations of precipitate sizes and volume fractions, that both aging and re-solution of precipitates occur in the nugget during welding of alloys 7075 and 2014. This is further supported by temperature measurements made during welding, which demonstrate that the theoretical dissolution temperatures for hardening precipitates in both 7075 and 2014 were exceeded. Su et al.[182] have argued that the temperatures achieved during the welding thermal cycle for alloy 7050-T651 were sufficiently high to take all of the strengthening precipitates in the nugget into solution. The cooling rate was not sufficiently rapid to prevent reprecipitation, but this was favoured by heterogeneous nucleation at dislocations and constituent particles, resulting in only large precipitates forming. It has been argued that the observed hardness peak in the nugget is due to a number of factors: a decrease in grain size, solid solution strengthening and comminution of constituent particles.[178,187] With regard to comminution, it has been postulated that some dissolution of constituent particles may occur as a result of the thermomechanical processing achieved in the stirred region.[178]

Precipitation studies in the HAZ of alloys 7075 and 2014 have shown full precipitation of the equilibrium phases,[6, 187,191] in common with arc welding.[235] Strangwood et al.[187] and Svensson and Karlsson[188] have shown that precipitation is favoured at matrix/ constituent particle interfaces. Full strength may be recovered only by a full solution anneal;[187] a post-weld aging heat treatment is not applicable.[6, 187] It is worth noting that some aging of the HAZ has been observed in alloys 2014A-T651 and 7075-T651,[178] suggesting that some reversion may take place, but this was within the first few hours after welding, after which no appreciable aging was observed.

Figure 26 shows 7050 welds made in 6⋅35 mm 7050-T7451 alloy. This temper is overaged, designed to give a good compromise between mechanical strength and corrosion performance. More recent TEM studies have systematically investigated the precipitation state in the HAZ, TMAZ and weld nugget, as a function of welding conditions (and also peak temperature of the thermal cycle, inferred from thermal modelling).[236] While TEM studies reveal precipitation behaviour in detail, they have limited quantitative capability, and are restricted practically to a few locations per weld. A powerful technique for obtaining quantitative data covering whole welds is small angle X-ray scattering (SAXS) using synchrotron X-rays. This is an expensive technique and not available routinely, but the studies conducted on FSW offer valuable insight into the process.[230,232,236]

Figure 27 shows representative SAXS maps for precipitate volume fraction and size of η precipitates in 7449 FSW. This technique was used to quantify and explain the differences in final precipitate distributions between T3 and T79 tempers, each welded at high and low speed. The SAXS data were supported by selective TEM observations and provide robust evidence for the interpretation of hardness profiles. In particular, the relative extent of coarsening in the HAZ, and dissolution in the nugget are clearly evident. The higher precipitate fraction in the plate in the T79 condition compared to T3 is also evident.

Fig.26. Comparison of microstructures in 7050 FSW[191] (TEM bright field images: welds made at 396 rev min-1, 102 mm min-1)

a) 7050 parent material: well characterised intragranular precipitates are η' [Mg(Zn,Cu,Al)2] and intergranular precipitates are η (MgZn2) and/or Mg3Zn3Al2;

b) area in HAZ, where same precipitates are found but thermal cycle has led to 56 increase in size;

c) area from recrystallised nugget, where some grains had significant dislocation densities: dislocations are pinned by Al3Zr dispersoids or Al7Cu2Fe inclusions

Fig.27. Small angle X-ray scattering maps of volume fraction and size of precipitates in friction stir welds of T3 and T79 7449 alloy produced at low welding speeds:230 FSW tool shoulder diameter 23 mm

Svensson and Karlsson[188] and Svensson et al.[237] studied the complex precipitation sequence in various regions of a friction stir weld in 6082, which is one of the most common alloys welded by this process. The normal hardening precipitate in this alloy is β'' (Mg5Si6), but this can dissolve easily at temperatures of 200-250°C, which are easily reached in the HAZ, and another precipitate, β' (Mg1⋅7Si), forms on dispersoids in the matrix very easily at ~300°C, a typical temperature in the HAZ. This is less effective as a hardening precipitate. However, in the nugget region, temperatures are much higher, allowing dissolution of all precipitates, and β' does not form easily, as the weld will cool very rapidly through the temperature range where such precipitation can occur. Figure 28 shows typical examples of HAZ and nugget microstructures.

Fig.28. Images (TEM) of friction stir welds in 6082-T6 alloy

a) in nugget zone dispersoids are free from precipitates;

b) in low hardness zone, needle-like precipitates of β'-Mg1⋅7Si are found on dispersoids[237]

Fonda et al.[196] have studied the precipitation sequence in an underaged 2195 alloy of 25 mm thickness. Figure 29 shows examples of the precipitate structure in the HAZ, TMAZ and nugget for 2195, and clearly exemplifies the significant changes which occur. Evidence is shown for the coarsening of the T1 and θ' precipitates. In the TMAZ, these precipitates are gradually replaced by a δ' phase (Al3Li), with none of the initial phases being present in the hottest part of the unrecrystallised TMAZ. Guinier-Preston zones were also detected at this point. Rod shaped precipitates, identified as TB phase (Al7Cu4Li), were detected in the nugget region. These often nucleate on β' (Al3Zr) which is difficult to distinguish from the δ' phase.

Fig.29. Transmission electron diffraction patterns and micrographs from HAZ, TMAZ and nugget region of friction stir weld in 2195 alloy ^[196]

Generally, the trends in precipitation observed in 2xxx and 7xxx series alloys have been matched in 6xxx series alloys. Coarsening of β' precipitates and solution of needle precipitates was observed in the HAZ of welds in alloy 6063.[142, 192,204,228] However, studies of precipitation behaviour in the weld nugget of 6xxx series alloys have produced some conflicting results. Murr et al.[8] published TEM micrographs of the weld nugget in alloy 6061 showing large precipitates present, but their identity was not established. However, Lienert and Grylls[189] reported that only second phase particles, presumably constituent particles, were present in the weld nugget of the same alloy; the hardening precipitate β'' was noted to be absent from the microstructure, implying that dissolution and not overaging had occurred during welding of this particular alloy. Temperature measurements taken during welding, which indicated that the precipitate dissolution temperature was exceeded, were reported to support this observation. A similar result has been reported in a series of studies on alloy 6063, welded in both the T4 and T5 conditions.[142, 192,204,228] Evidently, welding conditions have an effect on the thermal cycle experienced by the nugget and this may explain the observations. As observed in other precipitation hardenable aluminium alloy systems, aging studies have shown that joint properties may be improved by post-weld heat treatment.[204]

Modelling of microstructure and hardness evolution in heat-treatable alloys

The evolution of microstructure and hardness in heat-treatable alloy FSW has been modelled in most detail, adapting methods developed for arc welding.[238-244] For the HAZ, the problem is purely thermal; for the TMAZ and nugget there is the potential added complexity of coupling between the deformation microstructure and precipitation. These microstructure models fall into two categories:

1. semi-empirical (with some physical basis), based on isothermal heat treatments and indirect calibration via hardness measurement, and able to predict hardness profiles across welds
2. physically based, using detailed thermodynamics and kinetics of phase transformations, calibrated on direct measurement of microstructural features, and able to predict hardness and strength, with the potential for extension to ductility, fracture toughness, fatigue and corrosion properties.

The semi-empirical methodology has been applied to FSWin 2000, 6000 and 7000 series alloys.[51,52, 121, 168, 245-249] This method uses the kinetic strength concept, outlined in the section on 'Deformation microstructure in weld nugget' above. Isothermal softening experiments are conducted, and the hardness scaled to the residual volume fraction of hardening precipitates. A simple kinetic model for dissolution is used to calibrate the temperature dependent dissolution time t*(T). Thermal cycles are converted to equivalent isothermal holds, and the final precipitate fraction and hardness predicted, including the extent of subsequent natural aging. It is currently limited to artificially aged tempers (T5, T6, or T7), for which dissolution is the dominant mechanism. Figure 30 shows predicted and measured hardness profiles, including predicted curves immediately after welding and after natural aging.[52]

Fig.30. Semi-empirical model predictions and measured hardness profiles in 20 mm thick friction stir weld in 7449-TAF, at six different depths through thickness (after Colegrove et al.) ^[248]

More sophisticated approaches to the evolution of precipitation in heat-treatable aluminium alloys have been developed and applied to FSW.^{[242-244,250,251]} Both these and the simple semi-empirical models have been incorporated into integrated modelling platforms for FSW, spanning heat generation, thermal modelling and prediction of microstructure and hardness profiles for heat-treatable aerospace alloys.^[248]

In these analyses, the evolution of the full size distribution of precipitates is modelled, capturing the competition between dissolution, coarsening and transformation from one phase to another. Extensive use is made of direct measurement of volume fractions and particle radii by SAXS and electron microscopy (TEM or FEG SEM) for calibration and validation of the model. The key ingredients of the physically based methodology are:

1. thermodynamic calculation of phase stability for both metastable and equilibrium precipitates, employing thermodynamic database software
2. classical isothermal nucleation, growth and coarsening theory, applied to thermal cycles.

More than one population of precipitates may be considered simultaneously, with the competition between phases and evolution of each phase determined by the instantaneous microstructural state and temperature. Figure 31 shows an example of the predicted evolution of various phases for different locations in a FSW of alloy AA7449-T7.^[248]

Fig.31. Predicted evolution of a) precipitate volume fraction and b) equivalent radius in AA7449-T7 friction stir weld, for three positions typical of nugget, TMAZ and HAZ (after Colegrove et al.) ^[248]

The models are complex, requiring expert users, and have a significant computational penalty, but the potential benefits are large. For example, FEG SEM is able to provide independent data for grain bulk and grain boundaries. This opens up the potential for modelling the effect of dislocation structures on precipitation within the TMAZ and nugget, including quench sensitivity effects (i.e. precipitation of non-hardening phases during the cooling part of the thermal cycle). Strength and hardness predictions can also be made at a more detailed level than in the semi-empirical approach, using the predicted volume fraction and average radius.^[252,253] In principle, the detailed description of the precipitate state (including distinctions between grain interiors and boundaries) can be used to predict more complex but industrially critical properties, such as ductility, fracture toughness, fatigue and corrosion. Some recent results for fracture toughness modelling are presented by Derry and Robson.^[254] Microstructural models of this type require careful calibration to data from thermodynamic computation and high resolution microscopy. They are primarily used to develop scientific understanding, but, suitably packaged, can offer industrial users with a tool to reduce the number of experimental trials in, for example, a new joint design, for previously calibrated alloys. But there remains the prospect in future of using such tools for alloy development, in which the friction stir weldability of a new alloy variant is considered earlier in the development programme, alongside the core mechanical and corrosion properties.

Dissimilar welds

In agreement with the distinct macroscale separation discussed in the section on 'Mixing across dissimilar welds', close examination of weld cross-sections commonly shows alternating bands of each alloy, often only micrometres wide. Larsson et al.^[49] have described EDS scans of Mg across two boundaries between 5083 and 6082 regions. The transition between the two Mg levels is located to within a few micrometres, suggesting only limited diffusion occurs. No evidence of regions with an intermediate composition can be seen. Many other examples exist which show the sharp transition, for example that shown in Fig.32, which shows X-ray maps from 2219-T87/7075-T6 and 5083-H321/6082-T6 dissimilar welds. In both cases the sharp boundaries between the phases are clearly evident. Hardness data show the expected behaviour in the HAZ and unrecrystallised TMAZ regions. However, in the nugget different behaviour is often observed. The hardness in the nugget can oscillate between the hardness levels expected for each of the two alloys as seen in Fig.33a, although this is not always the case: in Fig.33b the response is simply a composite of the responses of the two constituent alloys (see Fig.22), with the 6082 hardness in the nugget immediately after welding being much less than after aging. In common with welds in a single alloy, the nugget region is not generally the origin of tensile failure, so in dissimilar alloys failure will be controlled by the HAZ/TMAZ of lowest strength in the two alloys concerned.

Fig.32. X-ray maps from nugget regions in dissimilar welds (100 µm markers)

a) Mg trace for joining 5083-H321 (light) to 6061-T6 (dark);

b) Cu trace for joining 2219-T87 (light) to 7075-T6 (dark)

Fig.33. Hardness traverses across 6mm thickness dissimilar welds:^[52]

a) 7075-T7351/2219-T6 and 2219/7075;

b) 6082-T6/5083-H321 ^[52]

In summary, both heat-treatable and non-heat-treatable alloys have marked variations in microstructure and hardness across FSWs. In the former the growth, dissolution and reprecipitation of strengthening precipitates across the welds must be accounted for, while loss of work hardening is important in the latter. Models capable of capturing these effects have been developed. For both alloy types the complex grain structure variations can be explained in terms of grain growth, recovery and recrystallisation driven primarily by the reduction in stored energy as a function of the local peak temperature attained.

Mechanical properties

There is an ever increasing volume of data in the literature on mechanical properties of friction stir welds.^[255] It is thus beyond the scope of the present review to summarise all aspects. An ISO standard (ISO 25239: Friction stir welding - aluminium) on FSW will be published in 2009; this uses other ISO standards to define standard mechanical tests. When considering specimen design, as for all welding processes, the dimension of the initial test specimens should have some relationship to the final structure, in particular to ensure that the heat sink is representative, and that a steady state is reached (at least approximately) and that the residual stresses are comparable. Finally it should be noted that hardness was discussed above as a useful means of delineating microstructural changes across friction stir welds.

Tensile properties

It is often stated that the tensile properties of friction stir welds generally equal or exceed those reported for fusion welds. Although this is often the case, some qualification of this statement is in order:

1. tensile properties of fusion welds made with a filler are often determined as much by the filler as by the welding process: in general when fusion welding aluminium alloys, the filler wire is not the same composition as the parent material, and therefore may not have the same mechanical properties
2. alloys such as 2xxx and 7xxx are designed to have high strength, and therefore the strength of welds is of particular importance. Unfortunately, these alloys are generally difficult, and sometimes impossible, to weld by fusion processes; thus, comparative data from high quality fusion welds is scarce, or non-existent, and comparisons with FSW are not always straightforward
3. when comparing tensile data on different types of weld, care should be taken to establish how the measurements have been made, e.g. removing the overfill in fusion welds may affect the properties, and not all authors state whether or not this has been done; determination of yield stress is dependent on the technique and equipment used (again not always stated).

When assessing tensile data, it should be remembered that, as shown above, the microstructure across a friction stir weld is typically highly non-uniform. As a result yield strength, tensile strength and ductility may change considerably over very short distances. Consequently very different results can be obtained according to whether the welds have been tested longitudinal or transverse to the weld. The stress-strain response will vary even for cross-weld tests according to the width of sample, since this will determine the retained residual stresses (see below), and the length of the testpieces since this will determine the average ductility/overall elongation, 0⋅2% yield stress, etc. There are many reports of low elongation in cross-weld tensile tests of welds; however, in many cases this is not due to low ductility, as confirmed by the significant reduction in area. Instead, the strain will have been concentrated in a very small part of the gauge length where a locally softer microstructure may have formed. Studies of deformation by Mahoney et al.^[6] on 6⋅35 mm 7075-T7541 and Liu et al.^[256] on 5 mm 1050-H24, 6061-T6 and 2017-T351 have demonstrated the variability in strain across transverse tensile samples. Consequently, overall elongation measurements made on cross-weld samples tend to be unrepresentative of any region of the weld and serve only to identify the likely failure location under static loading.

Two strategies have been developed for extracting more representative data to map the properties across friction stir welds, the former more suited to extracting longitudinal properties, the latter capable of extracting transverse properties:

1. the excision of matchstick style microtensile test samples: in this manner samples can be removed parallel to the welding direction x that are representative of parent, TMAZ, HAZ or weld nugget
2. the cutting out of thin cross-weld testpieces: here, the microstructure varies along the length y of the testpiece, but not through the thickness z or width x, provided the testpiece is sufficiently thin. In such a case deformation behaviour will vary as a function of y position. This can be monitored by full field strain mapping using laser speckle interferometry or digital image correlation, for example.

Microtensile testing has provided the bulk of the data delineating the variation in mechanical properties across the various microstructural zones of friction stir welds (Fig.34a and b). Such studies include those by von Strombeck and co-workers,^[257-259] Allehaux et al.^[260] using 10 mm thick 7349-T6 alloy and Denquin et al.^[261] using 6 mm thick 6056 alloy. In the last case, minimum ductility was reported in the centre of the nugget region; the stress-strain response parallel to the welding direction was measured as a function of lateral distance from the weld, e.g. Fig.34b.

Fig.34. a) hardness profiles for Al2024 friction stir weld at three depths,^[259]

b) corresponding longitudinal tensile performance as determined by microtensile specimens,^[259]

c) variation in hardness (bold circles) and 0⋅2% proof stress (open circles) as determined from cross-weld tensile test^[180] monitored by electronic speckle pattern interferometry for FSW AA5083 welded at 200 mm min^-1 and

d) corresponding evolution of tensile strain with position across weld as crossweld load is raised^[180] (HAZ boundaries marked by dashed lines)

Cross-weld testing can provide useful insights if the strain is measured as a function of position through the microstructural zones. Initially, this was carried out by monitoring closely spaced parallel lines^[6] or Vickers indents,^[204] laser extensometry^[262] or a small number of strain gauges.^[263] Mahoney et al.^[6] recorded the distribution of strain at failure for 7075-T651 alloy FSW which showed very close agreement with the hardness variation characteristic of FSW for the alloy. The peak elongation (~15%) corresponded to the HAZ with the strain in the weld nugget close to that of the parent in accord with their similar hardness. A more sophisticated approach is to measure the strains by digital image correlation^[264] or by electronic speckle pattern interferometry.^{[180, 265]} The variation in 0⋅2% proof stress across an AA5083 FSW joint are shown in Fig.34c. It is clear from Fig.34d that failure occurs just outside the tool shoulder because this is the softest region of the weld zone. In this case the performance perpendicular to the weld direction is measured as a function of lateral position from the weld, e.g. Fig.34d.

When collecting tensile strength data on the basis of cross-weld testing it should be remembered that friction stir welds typically have significant residual stresses (see below). Test samples cut from larger plates may retain a significant proportion of the weld stresses which may compromise tensile strength measurements. Only when the cross-weld sample width is less than the size of the tensile zone (approximately the width of the HAZ) is the residual stress negligibly small.^[266]

As a result of an international collaborative effort (Eurostir),^[267] data for a variety of alloys have been published: the performance of different alloy types and tempers from this and numerous other sources (where indicated) can be grouped together. Their generic hardness responses are shown schematically in Fig.23 and their mechanical performance is as follows.

For heat-treatable alloys (e.g. 2xxx, 6xxx or 7xxx alloys in the T6 or T7 condition), irrespective of the heat treatment condition, cross-weld tensile tests normally fail at the side of the nugget, at or close to the HAZ/TMAZ boundary.^[268] The failure mechanism is a ductile shear failure, showing 45° facets. In thicker samples, the faceting may be more complex, but the mechanism is the same. The elongation is almost invariably less than found in the parent material, due entirely to concentration of strain in softer regions. The local ductility at failure can be estimated from the reduction in area, and substantial necking, indicating good ductility, is common.

Tensile strength is typically about the same as found in the parent material in the annealed condition, although significant improvements can be made by minimising the thermal cycle. Joint efficiencies exceeding 90% have been reported in 7xxx alloys. Liu et al.^[269] reported joint efficiencies as high as 82% for 2017-T351. Sato and Kokawa^[204] report an inverted top hat profile for yield strength, the trough in the nugget and HAZ being 50% of the parent T5 condition for 6063Al. Upon post-weld aging 95% of parent strength in the weld region was recovered, although the fracture location remained unchanged.

Failures can occur on advancing and retreating sides, although for a series of welds they will usually all fail on one side or all fail on the other.

For work hardened non-heat-treatable alloys (e.g. 5xxx alloys in the H1xx, H2xx or H3xx conditions), failure of cross-weld tensile specimens normally occurs in the centre of the weld, where the hardness is at a minimum (Fig.34). For example, in Fig.15, the AA5454 joint has failed at a kissing bond.^[156] Despite this, the failure stress and elongation is only some 10% less than the parent. For AA1050-H24 failure was on the advancing or retreating side with the distance from the weld centre decreasing with decreasing pitch in correspondence with a narrowing troughs in hardness.^[268] The failure stress observed is typically close to the annealed strength of the material, although higher values can be obtained if the heat input is minimised. Elongation values are normally a little below the parent value, but the reduction is less than in heat-treatable alloys. Failures are fully ductile, with extensive necking. For annealed non-heat-treatable alloys (e.g 5xxx in the O condition), failure of cross-weld tensile alloys can occur anywhere on the sample, although they usually occur away from the weld in the parent material. Elongation is therefore typically the same as the parent material, and the failure mode is invariably very ductile. Thus joint efficiencies of 100% can be obtained.

One advantage often quoted for tensile properties in friction stir welds is the very consistent performance from weld to weld. This is perhaps well illustrated by data from Lockheed Martin,^[270] where the analysis of a large number of welds in a 2xxx alloy showed that FSW gave rise to a small increase in average tensile strength, but a very much reduced scatter band. As design of the component in question was based on minimum strength which can be guaranteed, the higher repeatability of friction stir welds allowed an extra 20% strength to be used in the design, even though the average strength of the friction stir welds was not much more than that of fusion welds. It should be noted that this is due to the low variability in weld properties rather than the incidence of defects. Typical data are shown in Table 3.

Table 3 Tensile data for variable polarity plasma arc (VPPA) and FSW joints showing benefit of FSW

Ideally, the loss of parent material strength in an FSW would be negligible. At present, parent material properties can only be achieved with annealed alloys which cannot be softened by further heating during welding. Since loss of strength and hardness is usually related to either overaging in precipitation hardened alloys, or annealing in work hardened alloys, minimising the heat input should offer a way of improving properties. However, this approach is limited by the fact that the material being welded must be hot enough to flow, and in aluminium alloys this temperature will cause softening.

Heat input can be minimised by several methods for example:

1. use of more efficient tool designs that require less energy to push them through the weld
2. use of higher welding speeds and/or lower rotation speeds
3. use of artificial cooling (water sprays, welding underwater, etc.)
4. active tool cooling.

Residual stresses

As welded residual stresses

Residual stresses in welds are of great significance in determining weld performance, in particular fatigue and fracture toughness. In fusion welding, residual stress levels are often at, or very close to, parent material or weld metal yield strength. In solid state welds the residual stresses can be substantially lower, although this is not necessarily so. In comparing residual stress levels in friction stir welds, great care must be taken in interpretation of the results, as there are several techniques which are commonly used. Determination of residual stresses is a complex area, and an authoritative summary of the methods which can be used is beyond the scope of this review. Readers are referred to other texts for more detailed information on the origins of weld residual stresses,^[271,272] as well as the measurement of residual stresses by destructive (e.g. hole drilling,^[273] contour method)^[274,275] and non-destructive (e.g. neutron diffraction,^{[273,276,277]} synchrotron X-ray diffraction^[277,278] and magnetic)^[275,279] techniques.

Several authors have determined residual stresses non-destructively using synchrotron X-ray diffraction. ^{[180, 280-284]} Although different materials were examined, there is broad agreement in the results, in that the longitudinal stresses tend to show the largest variation, being most tensile in the HAZ, lower in the nugget and compressive in the parent plate (Fig.35).

Fig.35. Longitudinal residual stress distribution normalised by pin shoulder diameter for friction stir welds in 7449,^[326] 2199,^[301] 6082,^[54] 2024,^[324] and 5083 alloy ^[54]

Similar trends have been recorded by Staron et al.^[285] using neutron diffraction. Destructive methods such as the contour method,^[286,287] the crack compliance method^[288] and incremental centre hole drilling^[289] have also been applied.

The characteristic magnitude and profile of the longitudinal stresses across a friction stir weld are shown in Fig.35 for a range of alloys. The longitudinal stresses are typically much greater than the transverse. As is clear from the figure, the stresses tend to be tensile over a region extending just beyond the diameter of the tool shoulder. The tensile region tends to encompass the nugget and TMAZ and reflects the extent of the hot region beneath the shoulder. The peak stresses are often found just inside or just beyond the shoulder radius. Often the peak stress lies within the HAZ despite the lower hardness often found there. Lower level compressive residual stresses are typically found in the parent plate beyond the HAZ. The depth of the tensile plateau below the tensile peaks and the presence of a subsidiary peak on the weld centreline appear to be alloy specific. It should also be noted that the breadth of the tensile region and the magnitude of the stresses vary greatly according to the processing conditions. For example, much lower stresses than those plotted in Fig.35 are reported for 2024,^[290] 5083,^[180] and 6013.^[281]

Although almost all the attention to date has focused on the variation of longitudinal stresses lateral to the weld line measured midthickness and midweld length, where the situation can be described as steady state (e.g. Fig.35), in practice the stress field varies both through thickness and along the welding direction. The variation through thickness for 20 mm 7449 plate is shown in the upper part of Fig.36. It is clear that the largest stresses are found near the surface (in this case directly under the edge of the shoulder). Further the stress field profile broadly follows that of the HAZ and the hardness variation being narrower at the base where the heat input is less. For plates less than 6 mm thick the stress field is essentially uniform with depth through thickness.^[282] The lower part of Fig.36 shows the inplane variation in residual stress for a 3 mm thick 6082/5083 weld. It is clear that, in the case of the 100 mm long weld shown, the largest stresses are not found midlength but continue to rise towards the end position. This is in contrast to TIG welding similarly sized plates where the stresses the fell towards the end of the plate as the plate warmed up and thus the mismatch between weld and plate decreased.^[277,291] Also, in contrast to TIG welding,^[291] the stresses around the start and end positions are not particularly high. Both these observations may be the result of lower heat input associated with FSW. The stresses are not symmetrical across the weld line in Fig.36 because the stresses for the 6082 are lower in the nugget than for the 5083,^[54] due to dissolution of the hardening precipitates there. The transverse stresses are generally much lower than those in the longitudinal direction with the most significant tensile stresses (about 50-100 MPa) found around the exit hole.

Fig.36. Longitudinal and transverse residual stress variation: above, plate cross-section in 20 mm thick AA7449/7449 friction stir weld (tool diameter 34 mm);^[326] below, at midthickness for dissimilar 6082/5083 weld (tool diameter 18 mm) in 3 mm thick plate^[54] (weld started at white spot and finished at black spot). In both cases advancing side is on right

If the stress is measured, as is often the case, on a cross-weld sample extracted from a larger welded plate, account should be taken of the possibility of stress relaxation. In essence a weld length approximately eight times the diameter of the tool must be retained if 90% of the residual stresses are to be retained on cutting out the testpiece.^[266,286] This criterion is often not fulfilled, which may explain the low stresses observed in some studies. ^{[281,286,292]}

To understand how the residual stresses arise, finite element modelling (FEM) provides a picture of how the stresses evolve during welding. In this respect it is important to note that current FEMs for predicting residual stresses tend to ignore the mechanical stirring effect of the tool and regard it simply as a heat source.^[293-296] Such models involve a decoupled heat transfer and a subsequent thermomechanical analysis. An attempt has been made to incorporate the mechanical effect of the tool using FEM, representing the down force and torque loading elastically, without incorporating large plastic strains.^[128] However, to take proper account of the material flow around the pin an arbitrary Lagrangian-Eulerian formulation must be applied to avoid excessively severe distortions of the mesh.^[152] Fluid dynamics based models are now beginning to emerge capable of predicting residual stresses, which explicitly take into account fluid flow.^{[119, 297,298]} Unlike purely thermal models, these are capable of predicting differences in stress between the advancing and retreating sides of the weld. Typically very little difference is either predicted or observed in practice; however, Fig.37 does show slightly high measured and predicted stresses on the advancing side and this observation is supported by the work of others.^[299]

Fig.37. Measured and predicted effect of traverse speed and rotation speed: experimental data are for 2199,^[301] predictions for 7050 ^[297]

The evolution of longitudinal stress as the tool passes is depicted in Fig.38. It demonstrates that ahead of the tool the compressive stress caused by the expanding hot material impinges on the compressive yield stress locus, causing local plastic straining. Just behind the tool longitudinal tensile stresses begin to generate as the weld material cools. Initially stress development near to the weld line is limited by the low tensile yield stress (Fig.38c). This local tensile plastic straining at the weld line results in the characteristic 'M' shape typically observed in the welded plate (Fig.38d), as the hot region plastically deformed in compression, to a greater width, ahead of the tool becomes stressed in tension as it cools behind the tool. As the tool travels forwards and the temperature falls, the tensile stress level builds up at a rate slower than that at which the yield stress rises so that a point is reached very soon after the tool has passed when no further yielding occurs and the increasing misfit is then accommodated elastically. As the tensile stresses develop near the weld line these are balanced by a compressive stress towards the edges of the plate. Lombard et al.^[300] have noted that for AA 5083 the width of the tensile region increases and the maximum tensile stress decreases with increasing heat input. In agreement with experiment,^{[180, 284,300,301]} the residual stress distribution has been found by modelling^[294,295] to be dependent on the welding traverse and rotational speeds (Fig.37). Despite the fact that the parametric study is for AA2199,^[301] and the modelling is for AA7050,^[297] the responses in Fig.37 show a high level of agreement. In both cases the 'M' profile characteristic of FSW shows a small secondary peak centred on the weld line, the peak situated just outside the tool shoulder is slightly higher on the advancing side and the width of the tensile region increases, and the magnitude of the tensile region decreases, as the traverse speed is lowered and the rotation speed is increased. This is believed to be because the high heat input associated with low traverse speeds and high rotation speeds (see Fig.5) leads to more extensive softening in the weld region. This in turn results in overall reduction in magnitude of the longitudinal residual stress, but a wider tensile stress zone (Fig.37). The residual stress field is governed primarily by the as-welded yield strength distribution in the stir zone, HAZ and TMAZ. On the other hand, the natural aging process, while strongly affecting the final weld strength, shows minimal influence on the residual stress. ^{[54, 297]}

Fig.38. Effect of external mechanical tensioning (given as percentage of parent plate yield stress) on predicted longitudinal residual stress profiles for AA2024-T6 plate friction stir welded at 770 rev min^-1 and 195 mm min^-1 (tensile and compressive yield loci are shown as dashed lines)^[296]

a) as heat source approaches and compressive stresses form;

b) directly through tool centre, showing resultant reduction in thermal strain by compressive yielding;

c) 8 mm behind centre of pin and at edge of shoulder, as heat source retreats, material begins to cool and tensile yielding occurs;

d) final stress state after removal of tensioning loads

Residual stress control

Residual stresses arise from the accumulation of misfit between the weld region and the remaining plate. There are a number of means by which the misfit, and hence the residual stresses, can be manipulated, e.g.

1. thermal tensioning
2. mechanical tensioning
3. subsequent processing treatments.

One of the earliest reported applications of thermal tempering was by Greene and Holzbaur,^[302] who in 1946 used superimposed temperature gradients to achieve reduced residual stresses in longitudinal butt welds in ship hull structures. Local induction heating has also been investigated for residual stress improvement.^[303] Michaleris and Sun^[304] and Dull et al.^[305] have applied thermal tensioning to reduce buckling distortion, whereas Dong et al.^[306] developed an in-process thermal stretching technique for effective mitigation of residual stresses and distortion on repair welding of aluminium panels. Barber et al.,^[307] van der Aa et al.^[308] and Williams and co-workers^{[285,309-311]} have applied local cooling to FSW, using either solid or liquid CO₂ trailing the heat source, as a means of creating dynamically controlled low residual stress and distortion free welds; others have used water jets.^[312]

Several mechanical tensioning systems have been proposed. Yang et al.^[313,314] have mechanically compressed the weld on cooling using a pair of rollers positioned on either side of the weld line, reducing both residual stress and buckling distortion. Williams and coworkers^{[79, 296,315]} have shown that the application of global, or far field, mechanical tensioning externally during the welding process can greatly reduce the tensile residual stresses in FSW joints. In global mechanical tensioning a load is applied uniformly along opposite ends of the plates prior to clamping the parts for welding, so that a uniform tensile stress is maintained in the two butted plates parallel to the weld line. The clamping and tensioning loads are then released after the FSW tool has traversed along the join line forming a weld. Perhaps counter intuitively, Williams et al.^[315] found that high levels of mechanical tensioning parallel to the welding direction can actually reverse the state of stress, so that compressive longitudinal residual stresses are found in the weld region.

Global mechanical tensioning operates by reducing the compressive bow wave ahead of the travelling heat source and increasing the tensile plastic strain developed in the hot zone trailing the weld. Figure 38 indicates that for low levels of tensioning (<40%) a reduction in the compressive plastic strain field ahead of the tool, as the hot material expands, is mainly responsible for reducing the final residual stresses. At higher tensioning loads, little or no compressive misfit develops ahead of the tool. Instead, larger tensile plastic straining of the softened hot material after the tool has passed causes a tensile misfit, or 'over tensioning', once the tensioning forces are removed. This results in the compressive longitudinal stresses seen along the weld line (Fig.39b). Taken together, these effects give rise to the observed approximately linear reduction in longitudinal weld stresses with tensioning level and zero residual stresses can be engineered at tensioning levels of about 40-50%, depending on the material and welding conditions. The results indicate that while the level of residual stresses present in the untensioned case is a function of the alloy, the rate of residual stress reduction brought about by mechanical tensioning is essentially alloy independent (Fig.39b). In all cases studied it is essentially linear with respect to the tensioning load, so that the tensioning required to reduce the weld stresses to zero can be calculated directly from the stresses present in the untensioned case. For thin plates a guideline rule is that 1 MPa of tensioning reduces the tensile stress by approximately 1 MPa. Global tensioning was found to be less effective at greater depths in thick plates. Furthermore a reduction of the bending distortion and an increase in angular distortion was observed with increased tensioning, while no effects on the weld microstructure and hardness were observed.

Fig.39. a) comparison of measured (left)^[266] and predicted (right) longitudinal stress profiles for AA7449-W51 welded plates as function of tensioning level (0, 5, 10, 20, 30% of parent alloy yield stress): dotted profile represents predicted untensioned (0%) case for which there were no measured results;

b) residual stress at midthickness near weld line as function of applied global tensioning level for various alloys ^[290,326]

Besides post-weld heat treatment,^[316] there are a number of post-weld treatments by which the misfit introduced during FSW can be reduced to lower the residual stress level, or replaced with another one leading to beneficial residual stresses. Burnishing,^[317,318] laser and conventional shot peening^{[263, 319-321]} have been used to introduce new misfits and thus modify the near surface state of friction stir welds - introducing compressive stresses that are beneficial to the fatigue and stress corrosion behaviour.^[319,322]

It is also possible to apply the roller tensioning method described above after welding.^[323] Whereas during welding two rollers are passed on either side of the weld, a single roller approximately equal in width to the FSW tool shoulder can be rolled along the weld line once the weld has cooled. Recent stress measurements suggest that this approach is much more effective than when applied during welding (Fig.40) with loads in excess of 15 kN leading to compressive weld stresses for 2199.^[324] This compares with little effect during welding using two rollers and a combined down force of 75 kN. By contrast post-weld mechanical tensioning is much less effective than that applied during welding.^[296]

Fig.40. Longitudinal midthickness residual stress profiles as function of post-weld roller tensioning load for AA2199 friction stir weld:^[324] roller is shown inset, its footprint is indicated by horizontal solid line

Distortion

There is of course a strong link between residual stresses and distortion in welds. In most cases, FSW of aluminium produces low distortion levels compared with arc welding. However, significant distortion can occur in friction stir welds, in particular in thin gauge welds where the design imparts an asymmetry in restraint or heat sink. Preliminary studies show that thermal and mechanical tensioning methods used to control stress are also effective at reducing longitudinal distortion.^{[79, 325,326]}

In summary, residual stresses and distortion in friction stir welds in aluminium alloys can be engineered to be considerably less than those typical of fusion welds. The characteristic tensile longitudinal stress tends to be of lower magnitude but broader in extent the greater the heat input. As a solid state welding process there is considerable scope for manipulating the state of stress during welding. Thermal and mechanical tensioning have been found to be successful in reducing and even reversing the state of stress. Roller tensioning after welding has also been found to be successful in lowering the tensile residual stresses.

Fatigue

As with other mechanical property data, care needs to be taken to ensure that full information on the test procedure is available. In particular, in some fatigue tests, the surface of friction stir welds is dressed to remove flash and the surface markings. Occasionally even more material is removed, and this is acceptable if the weld in question is machined in the same way in service. When comparing data, R values should be checked to ensure they are the same especially in cases where residual stresses may exist in the testpiece, but these are not always reported. Examination of the literature has shown that the following generalisations can be made:

1. in simple S-N tests on cross-weld samples, the fatigue performance of friction stir butt welds is typically less good than that of the parent material tested under the same conditions.^[327,328] Many studies have found that after milling the top surface, the fatigue performance of 2014, 6013 and 7475 FSW joints approached that of the parent alloys,^[329-331] yet in other studies^[332] the properties remain significantly below parent material benchmarks
2. the fatigue performance of friction stir butt welds generally comfortably exceeds that of comparable fusion welds,^{[93, 226, 330,331,333,334]} a trend reported for many alloy grades
3. failure is normally (but not always) associated with an initiation event at the geometric stress concentration at the side of the weld on the upper surface;^[355] where this has been machined away, failure normally initiates in the region of lowest strength. For many alloy groups, these two locations are very close together
4. residual stresses can play a significant role in the fatigue behaviour191,336-338 and vary according to crack test geometry.^[339] In this respect it should be noted that the dimensions of most crack test specimens lie between the limits for negligible and complete relaxation of the residual stresses introduced by FSW.^[266]

In general, for machined welds, for which surface finish is not an important influence, three factors appear to play a dominant role in the fatigue performance: residual stress, microstructure and defects. As a consequence, without a detailed picture of all three and an understanding of their interactions, it is difficult to discern simple trends as a function of FSW processing conditions from the literature. Indeed, because of the overriding and interacting importance of these three poorly reported factors much of the data appears inconsistent and contradictory.

Dalle Donne et al.^[338] found that residual stress was by far the most important effect in their compact tension fatigue study in which fatigue crack propagation varied strongly as a function of R (maximum/minimum stress ratio). In agreement with others^[340,341] they found that once residual stress was taken into account very little variation in fatigue performance was observed as a function of FSW process parameters.

Naturally, as the crack grows the residual stresses redistribute. This can be especially important in considering residual stress effects on crack growth across the weld line (i.e. transverse) because the longitudinal stresses are typically the largest (see above). A number of researchers have calculated the effect of residual stress redistribution on the stress intensity factor for cracks growing through FSW joints.^[299,342] Figure 41 shows the initial residual stress field in an FSW testpiece, as determined by the crack compliance technique, along with the contribution to the stress intensity arising from the residual stresses.^[299] It is clear that the contribution can be very significant.

Fig.41. Residual stresses in 3⋅2 mm thick 2024 friction stir weld testpiece, measured by crack compliance technique (horizontal dashed line indicates compressive yield strength of material)[299]
a) initial residual stress before crack (solid line) and contribution of crack to stress intensity factor (dashed line) as function of crack growth for transverse crack;
b) corresponding residual stress redistribution ahead of crack tip

In general the refined microstructure characteristic of FSWs leads to improved fatigue properties compared with fusion welds.[226, 331,343]

Unsurprisingly, the presence of certain defects, especially root flaws such as zig zag kissing bonds containing oxide defects, can compromise fatigue performance.[161, 343-347] James et al.[348,349] comment on planar onion ring defects associated with the weld nugget that can accelerate the link-up of cracks in 5083 and 5383. Dickerson and Przydatek[350] suggest that root flaws up to 0⋅35 mm deep are tolerable without a significant loss in performance when compared to nominally flaw free welds. Widener et al.[351] found that for FSW 2024 once defects were eliminated by careful process control similar fatigue lives were obtained across a wide range of process parameters.

Effect of alloy system

The plethora of fatigue studies can be broken down by alloy system.

Extensive research has been carried out on 2xxx alloys.[343,346,351-353] For example, Fig.42 compares compact tensile crack propagation tests through welded and parent (TL: transverse-longitudinal section) 2024-T351 material (cracks parallel to the weld direction). The lowest threshold ΔK values were found for cracks propagating in the HAZ, which corresponds to a region of minimum hardness. By contrast, cracks propagating through the weld grew at rates 10 times slower than those of the parent plate. The largest threshold ΔK values corresponded to the TMAZ, which was some 15 times that for the parent plate. All the data appear to converge at large ΔK. To investigate the origin of these differences, the residual stresses were relieved by over 90% by a 2% stretch normal to the weld direction, which was insufficient to appreciably affect the microstructures. The fact that the fatigue crack growth rates for weld line and TMAZ material subsequently coincide with those for the parent plate was taken as evidence that the differences in crack growth rate were primarily due to residual stress. After stretching, the crack growth rate in the HAZ material remains inferior to parent and weld metal, which may be a microstructural effect. For thumbnail cracks growing across the welds, the opposite effect was observed with the slowest rates corresponding to the HAZ and the fastest in the TMAZ. This is consistent with the hypothesis that the residual stresses control fatigue crack growth resistance since the longitudinal stresses in 2024 are normally significantly tensile, with the TMAZ having the largest tensile stresses and the HAZ the compressive or low tensile stresses (see Fig.35). Milan et al.[354] found that under transverse cracking compressive residual stresses increase fatigue resistance in 2024 until the more brittle weld region is reached.

Fig.42. Crack growth data for FSW in 2024-T351 for cracks growing parallel to weld in CT samples:^[341] plots are for cracks located at various distances from weld line propagating parallel to weld a) in as-welded condition;

b) after 2% stretching (filled and open symbols correspond to different samples)

A number of studies have been undertaken on 5xxx series alloys.^{[336,337,355-359]} James et al.^[348,349] found that lower welding power (but sufficient to give 'adequate' plasticisation for a good weld bond to be made) gives the highest observed fatigue lives. However they found little correlation between the static and dynamic performance indicators, suggesting that crack path effects associated with onion skin defects are the primary cause of fatigue crack initiation and growth.

Most of the work concerning the fatigue properties of 6xxx series alloys focuses on 6055,^[357] 6056,^[316,360] 6061,^{[334,336,361]} and 6082.^{[316, 333,334,340,357,362,363]} It has been found that fatigue crack growth rates for compact tension specimens in the dynamically recrystallised zone are lower than for the parent metal especially at low DK. As in other systems this has been ascribed to the beneficial effects of compressive residual stresses in competition with detrimental grain refinement that brings about intergranular failure.^[361]

Some fatigue crack growth data have been reported for 7xxx alloys.^{[191, 339,353]} Jata et al.^[191] carried out eccentrically (or extended) loaded single edge tension tests on 7050-T7451 with the crack running through different regions parallel to the weld direction. They found that, at a load ratio of 0⋅33, fatigue crack propagation in the weld nugget region is inferior, while the growth rate in the HAZ is superior to that of the parent (Fig.43). On the basis of crack closure, fractography, microstructure and residual stresses they concluded that compressive residual stresses dominate fatigue crack growth in the HAZ, whereas for the fine (1-5µm) grain dynamically recrystallised weld nugget region, microstructure and intergranular failure mechanism dominate. This observation is corroborated by a tendency for differences in fatigue crack growth rates to become smaller at a load ratio of 0⋅7 (Fig.43). Kumagai et al.^[364] found that for 7050 the fatigue strength was close to that of the base metal with fracture in the softer HAZ.

Fig.43. Comparison of fatigue crack growth rates in laboratory air between weld nugget and HAZ for friction stir welds in 7050-T7451 plate at load ratios R=0⋅33 and 0⋅7: specimens were in as FSW+T6 condition ^[191]

Lap welds

There is much less information available on lap welds, and comparison with welds made by other processes is not possible at present. However, recent work by Ericsson and Sandström^[365] has shown that improved tool designs which increase the volume of disturbed metal at the interface, and which minimise the plate thinning and hooking defects associated with inadequately optimised lap welding can lead to significant improvements. Similarly, recent work by Thomas et al.^[366] has reported promising results for lap welds in 5083-H111 made using an advanced tool design. Friction stir welded lap joints in 2024-T351,^[367] and 7075-T7351,^[368] have shown improved fatigue performance over mechanically fastened joints of similar geometry when tested under similar conditions. However, Shepherd has indicated that in 2024-T351 lap welds, an improvement over bolted joints is only obtained after weld surface removal.^[369] Christner et al.^[92] have published information on fatigue of lap welded joints in thin gauge 2024-T351, and found that fatigue properties equalled or exceeded those of bolted joints. There is substantial evidence that more complex tool motions may also be useful in improving lap weld quality.^[370] As discussed above, single lap joints inevitably contain two crack-like regions, which can be straight or deflected (termed hooking, see Fig.18). Fersini^[371] has considered these regions as cracks in order to estimate the fatigue lifetime using FEM. The lifetime of the specimen is the time necessary for one of these cracks to propagate up to a critical length. Fatigue efficiency of some 15% of the static strength was estimated, the cracks failing under a mixture of mode I and mode II.

Fatigue extension strategies

Until recently, little attention has been paid to fatigue improvement techniques. Most of the attention has been focused on the use of mechanical surface treatments designed to place the near surface region into a state of residual compression. Laser and shot peening have been examined.^{[263, 319-321]} Typically shot peening introduces compressive residual stresses to a depth of around 200 µm, but laser peening can introduce residual stresses to a depth in excess of 1 mm. In some cases shot peening has been found to be effective,^[319,321] whereas in others^[263] only laser peening provided a significant increase (~120% compared with 10% for shot peening) in fatigue life. Low plasticity burnishing^{[317,322, 372]} can also introduce deep (>1 mm) compressive residual stresses. Jayaraman et al.^[322] report increases of 80% in the high cycle fatigue endurance of aluminium alloy friction stir welds.

In summary, although good fatigue results are consistently obtained with FSW, it is premature for design codes and joint classifications to be relaxed for friction stir welds, although this may in due course be justified. Lomolino et al. have collected and statistically analysed fatigue data on FSW for a range of alloys to derive a first set of reference fatigue curves.^[373]

Fracture toughness

Fracture toughness is not normally a problem in aluminium alloys, but there are nevertheless areas where it is important, in particular at very low temperatures such as those encountered in cryogenic structures, or in exterior surfaces of airframes. Fracture toughness has been studied using a number of testing configurations for alloys including 2014,^[374,375] 2024,^{[257, 376]} 2139,^{[116, 377]} 2195,^[378]378 5083,^[374,379] 6061,^[257,258] 7075,^[374] and 7449.^{[254, 380]}.

Mochizuki et al.^[379] found that for 5083-O, in contrast to hardness and static strength, the Charpy impact energy and critical crack tip opening displacement (CTOD) in the friction stir weld are much higher than those corresponding to the parent metal or the HAZ. This was ascribed to the fact that the fine grained microstructure in the stir zone helps to increase ductile crack initiation and propagation resistance.

Dawes et al.^[374] investigated the R curve behaviour of 2014A, 5083 and 7075; typical data are shown in Fig.44. This shows that the fracture toughness in the nugget and the HAZ/TMAZ region exceeded that of the parent material, presumably because of the very fine grain structures. This effect was found for all alloys tested, although the magnitude of the difference varied between alloy types. von Strombeck et al.^[259] obtained similar results on 2024-T351, 5005-H14 and 6061-T6 alloys using a CTOD parameter to measure toughness rather than the J parameter used by Dawes. Only 2024 joints exhibited similar or slightly lower fracture toughness than the parent alloy. This behaviour was attributed to changes in the characteristics of the inclusion and precipitates population. Supporting these observations, Brinkmann et al.^[258] have reported that the fracture toughness of the nugget in 3 mm 6061-T6 friction stir welds (and repair welds) exceeds that of the parent material by a considerable margin.

A number of studies on the fracture toughness of 2195-T8 have given broadly similar results.^{[377,378,382]} Kroninger and Reynolds in particular studied R curve behaviour in 2195.^[378] The observed improvement in R curve behaviour for weld line and nugget seen in Fig.45 is ascribed primarily to microstructural origins: these regions being solution treated and naturally aged by the tool giving greater toughness than the overaged HAZ and peak aged parent. The poor response of the VPPA weld is ascribed to brittle solidification products absent in the FSW.

Fig.44. Toughness data for 2014-T6 friction stir welds ^[381]

Fig.45. Comparison of R curves for cracks growing through parent plate, weld centreline, weld nugget and tool shoulder regions of 2195 friction stir weld, and for VPPA weld centreline ^[378]

More recently Kristensen et al.[267] published limited data on 6 mm thick 2024-T3, 5083-H111 and 6082-T6 alloys. Conventional CTOD tests at room temperature showed that the fracture toughness of the nugget in all cases equalled (for 2024-T3), or exceeded, the parent material values, in contrast to the R curve behaviour reported by von Strombeck et al.[259] Mochizuki et al.[379] studied the toughness of welds in 25 mm thick 5083-O at ambient temperatures and at -196°C. At both temperatures, the toughness of the stir zone exceeded the toughness of the parent plate by a substantial margin. HAZ toughness was slightly above that of the parent material toughness.

A good example of the importance of microstructure for toughness is given by the work of Derry and Robson[254, 380] on FSWs in 7449-TAF. Failure is predominantly intergranular, with a low level of transgranular failure through microvoid coalescence. This is because the metastable η' precipitates are semicoherent with the matrix and therefore allow dislocations to pass through them, resulting in inhomogeneous slip and stress concentrations where slip bands meet the grain boundaries. Grain boundary failure is exacerbated by low strength precipitate free zones at the boundaries. In the HAZ, precipitate coarsening and transformation to incoherent equilibrium η precipitates means that dislocations are held up, leading to more homogeneous deformation. Thus failure occurs by conventional nucleation and growth of voids at coarse intermetallics rather than at grain boundaries, and consequently toughness is higher. In the weld nugget, temperatures are sufficient to cause precipitate dissolution (Fig.27) and subsequently natural aging on cooling. The resulting fine dispersion of Gunier-Preston zones and g9 precipitates leads to intergranular failure in the nugget by the same mechanism as in the parent material. In fact, the situation is worsened by the increase in grain boundary area and the coarsened grain boundary particles mean failure is wholly intergranular. As a consequence the HAZ is eight times tougher than the parent and the nugget half as tough, as determined by Kahn tear testing. These effects can be predicted for the HAZ in terms of simple hardness[383] models simply by assuming toughness to be inversely proportional to hardness.

In summary, all the results confirm that microstructural factors play a determining role in fracture toughness. Typically the FSW nugget zone shows a higher toughness than the parent alloys. This is in contrast to the fatigue behaviour where residual stresses dominate, presumably because the plastic strain washes out any stored residual stresses. It is noteworthy that as well as the expected correlation between increasing hardness and strength and decreasing fracture toughness across FSW in three alloys systems, Dawes has also found remarkably good correlation between Charpy data and more sophisticated fracture mechanics J integral test methods[374] (Fig.46).

Fig.46. Charpy energy-toughness correlations for friction stir welds in various alloys ^[381]

Corrosion

It is well established that microstructure is an important factor in determining the corrosion behaviour of aluminium alloys.^[384] A great deal of attention has been focused on the Cu containing 2xxx and 7xxx series (e.g. 2024,^{[48, 385-390]} 7010, 7050, 7075),^{[385,388,391-393]} which show that the nugget becomes sensitised (Table 4). The severe themomechanical processing refines the grain structure and alters the precipitation distribution and chemistry, particularly near the grain boundaries (Tables 4 and 5). The relationship between microstructure and corrosion for 2024 is summarised in the form of a time-temperature-corrosion map^[384,410] in Fig.47. From this it is clear that the maximum thermal excursion in the FSW weld region is above the knee in the time-temperature-corrosion curve and the cooling rate sufficiently slow for pitting and intergranular corrosion to occur.

Table 4 Summary of corrosion observations for FSW Al alloys

Table 5 Typical relationship between microstructure, significant localised corrosion mechanisms and main investigation techniques for high strength aluminium alloy friction stir welds ^[414]

Fig.47. Time-temperature-corrosion plot for 2024 friction stir weld based on interrupted quench experiments, showing predominant corrosion mechanisms determined on basis of accelerated corrosion tests on sheet:410 time-temperature band added by Lumsden^[384] to indicate conditions typical of most sensitised FSW HAZ region

Localised corrosion

Pitting, the removal of metal at localised sites, is a common corrosion mechanism in FSW Al alloys. The tendency for pitting is characterised by the pitting potential, namely, the potential at which the protective anodic film breaks down. Connolly et al.[395] and Paglia et al.[407] have used micro-electronic cells to map this spatially. The former team used a micro-electrochemical test set-up (droplet cell) comprising a three electrode cell within a fine (10 µm to 1 mm) pipette mounted on an optical microscope for precise positioning to delineate the variation in pitting potential as a function of position across the microstructural zones. For 5456 the breakdown potential for parent, HAZ and TMAZ were similar; however, the nugget was significantly lower in accordance with the higher level of corrosion attack found there. This occurs due to preferential attack of β-phase precipitates. Interestingly the advancing side of the nugget was found to corrode preferentially. For the 2024 welds discussed above the pitting potential (Fig.48) was found to vary with increasing heat input,[396] again in accordance with Fig.47. At low heat inputs the whole TMAZ is most affected, but with increasing heat input the HAZ becomes significantly more affected than either the parent or nugget. Gel visualisation comprising agar, NaCl and universal indicator can also be used to highlight oxygen reduction (blue/green) and hydrolysis of metal ions (yellow).[395] It can be seen in Fig.49 that this method is in good agreement with the pitting potential results, showing increased corrosion in the HAZ as the heat input is increased, the nugget becoming less affected.

Fig.48. Pitting breakdown potential measured using droplet cell with NaCl across five 2024 FSWs described in Fig.49 ^[396]

Fig.49. Macrographs (left) and corresponding gel visualisations (right) of corrosion attack on cross-sections of 2024 friction stir welds after immersion in 0⋅1M NaCl for 24 h: welds were produced at different speeds and are arranged from top to bottom in order of decreasing heat input ^[396]

There is a general consensus that for copper containing alloys (e.g. 2024, 7xxx) intergranular attack (see Table 4) is encouraged by the precipitation of a network of Cu rich intermetallics at grain boundaries (Fig.50b), which lead to locally depleted zones that are vulnerable to localised attack,^{[391,394,395,411]} This vulnerability has been linked to the lowering of the pitting potential (Fig.48). For 7xxx and 2xxx alloys intergranular corrosion is thus generally preferentially observed in the TMAZ/nugget and HAZ depending on the heat input (Table 4). Figure 50a shows a scanning electron micrograph of a 2024 weld following immersion in Exco test solution for 8 h, revealing a severely corroded, narrow band in the HAZ. On the right side of the corrosion band, i.e. in the parent alloy, little development of corrosion is observed. Conversely, on the left side of the corrosion band, i.e. within the TMAZ, a number of randomly distributed pits, of dimension 50-300 µm had developed. Because 7075 contains less Cu than 7050, the enrichment of Mg and depletion of Cu is less, leading to lower susceptibility.^[412] For other alloy systems, for example 2195, 2219, 5983, 5456, 6061, 6081, the proclivity for the weld region to corrode is much reduced being more similar to that of the parent (Table 4). Hu and Melekis^[398] ascribe the better pitting resistance of 2219 compared with 2195 to the lower Cu concentration. Generally the welded zones for 2219 and 2195 show more non-uniform pitting, but better corrosion behaviour than the parent alloys.

Exfoliation occurs where corrosion products having a larger volume than the metal they consume produce a wedge-like action.^[384] It is not common in the weld nugget because large flat grains are much more susceptible than the fine grains typical of FSW. Exfoliation corrosion is usually tested by exposure to an oxidising acidic chloride (Exco) solution using ASTM G34.

In summary, corrosion of FSWs in 2xxx and 7xxx alloys continues to be a challenge; those interested are directed to more comprehensive reviews of the subject.^[413,414] It should be borne in mind that FSWs will naturally be associated with residual stresses; these may well play a role in corrosion studies. Finally, some work has been undertaken on the corrosion performance of dissimilar systems.^{[388,415-418]} In such cases there is clearly an opportunity for galvanic corrosion in the nugget, for example 2024 is anodic with respect to 7010. ^[388]

Fig.50. Scanning electron micrographs of intergranular attack on 2024 friction stir weld^[394]

a) general view following corrosion testing in Exco solution for 8 h;

b) network of CuMgAl2;

c) close-up of intergranular corrosion

Stress corrosion

The location for failure by stress corrosion cracking or corrosion fatigue is often the HAZ of 7075 and 2049 alloys or the TMAZ/HAZ for 7050 and 5454 (Table 4). If the weld is not particularly susceptible, as for 2195 and 2219 (Table 4), failure typically occurs in the softest region of the weld. Intergranular failures are usually associated with the sensitisation of the microstructure. Increased pitting corrosion may act as an initial stage for intergranular corrosion.[414] In constant strain rate tests, FSWs in 7075 have been found to exhibit much better environmental cracking resistance than 7050,[404,419] failure occurring in the HAZ in accordance with the increased susceptibility there (Table 4).

Strategies to reduce corrosion susceptibility

Short term post-weld heat treatments (artificial aging) (PWHT/PWAA), with thermal exposures similar to that during welding, may be an efficient way to rehomogenise the sensitised microstructure and thereby increase the corrosion resistance of the welds.[414] Such treatments may also reduce the level of residual stresses, thereby lowering susceptibility to stress corrosion. Lumsden et al.[419] found that 7050 is very sensitive to stress corrosion, exhibiting a ductility just 13% of the in air ductility when slow strain rate tested (according to ASTM G129)[384] in NaCl solution. Aging for one week at 100°C (equivalent to 10 years natural aging) restored the ductility to 80% of the in air value. Alloy 7075 was found to be less sensitive to SCC (73% of strain to failure in air), with a T73 post-weld temper restoring this to 85%. Widener et al.[420] found joining 7075 material originally in the T73 condition followed by PWAA to be preferable (in terms of higher tensile and yield strengths and better exfoliation corrosion resistance) to welding in the T6 temper condition followed by aging to T73. Retrogression and re-aging treatments were not found to improve joint properties of T6 material due to the severity of the overaging in the HAZ caused by the welding process. Merati et al.,[421] however, did have success with a localised retrogression and re-aging treatment to 7475-T73. Overall, 4 h at 190°C was found both to stabilise the microstructure and to enhance the corrosion resistance for 7075, with only a slight reduction in tensile strength and the added advantage of annealing out residual stresses.420 Less precise treatments have been applied using local heating of joints with torch flames in 7075 and 2219.[414].

The effect of starting temper of the parent material and PWHTs were investigated for FSW joints in 2024.[422] It was found that the exfoliation resistance of FSW Al 2024 joints may be restored through PWHT to the T81 temper or when initially welded in the T81 temper, followed by naturally aging.

Other approaches to reduce the sensitivity to corrosion include modifications to the tool design; for example, by using a tool with a scroll shoulder instead of a threaded pin/flat shoulder tool when welding 7050, the sensitivity to SCC can be eliminated.[423] Other studies have looked at the minimisation of corrosion through optimisation of the welding procedure for 6xxx alloys.[424]

The effect of chromate, molybdate and cerium nitrate inhibitor additions to NaCl solution were examined for dissimilar 7075/6056 FSW joints.[425] The chromates were better in terms of inhibition efficiency and inhibit all regions of the weld area to a similar extent. Though less effective, from an environmental viewpoint, molybdate and cerate may offer advantages.

Approximately 30 µm micro-arc oxidation coatings have been successfully applied to 2219 and 7018 FSW joints subjected to salt spray corrosion.[426]

Other surface treatments shown to be successful include low plasticity burnishing to retard corrosion fatigue of 2219 FSW joints.[372] An 80% increase in corrosion fatigue performance has been achieved for the same system by LPB,[322, 427] completely mitigating pitting corrosion damage, with comparable fatigue perfrmance regardless of salt fog exposure. Laser surface melting has also been found to be effective in reducing corrosion of 2024.[395]

Comparison with other joining processes

A reasonable volume of data (e.g. Table 3) exists to compare the properties of friction stir welds with other processes. Consequently, it is now well established that the mechanical properties associated with FSW joints are generally better than those obtainable in arc welds. However, mechanical properties are only part of the picture: process economics are also of significance, as is the quality of the weld which can be reliably obtained. For example, Hori et al.[330] have described work on a Japanese alloy (6N01) similar to 6005. Comparison of tensile properties with MIG and laser welding showed no improvement in tensile properties over laser or MIG (in fact FSW performed slightly worse), but a significant improvement in fatigue performance was reported. Furthermore, the tolerance on misfit, the remarkably low weld to weld variability and freedom from defects made the FSW process more attractive than its competition. A further example can be found in a recent study by Gesto et al.,[84] who compared FSW and GMAW of 6082-T6 for a marine application. They concluded that FSW is superior in terms of properties and weld quality, but at the price of higher capital costs.

In summary, in many cases improved mechanical properties are only part of a complex decision making process for selecting FSW. For example, the low defect incidence and low repair rates have been instrumental in process selection, especially for critical applications such as rocket fuel tanks.[90,91] The reverse argument is also true. In a number of cases companies have judged the potential advantages of the process in terms of mechanical properties and weld quality to be outweighed by the capital costs and other process conversion costs, particularly in situations where an expensive machine could not be kept fully occupied, or in areas where low cost is the highest priority.

Concluding remarks

Friction stir welding of aluminium is now a mature and robust process, which is becoming increasingly well established in the fabrication of critical components. It is true to say that FSW has extended the use of welding in certain materials and applications, in particular in the welding of 2xxx and 7xxx alloys for the aerospace industry. The qualities making the process attractive include reduced cost, minimal repair requirement, good properties and total automation leading to a high level of consistency.

At the present time, FSW can compete with other welding processes for quality of welds and performance. It should be noted that FSW is still relatively new, and has been in commercial production for less than 15 years. Nevertheless, progress has been significant, and further improvements and developments can be expected.

As with fusion welding, FSW is basically a thermal process. Temperatures reached (typically around 500°C) are sufficient to cause major microstructural changes in precipitation hardened or work hardened alloys. Unlike fusion processes, FSW also involves extremely high shear strains and strain rates, which will have a profound influence on the development of microstructures. The debate on the relative importance of recovery and recrystallisation continues to be lively, and further work is required in this area to gain a full understanding of the complex processes and their interactions which determine microstructures.

Friction stir welding is already one of the most energy efficient processes available, although improved process developments (in particular better tool designs) will no doubt further reduce the energy required to make the weld. Now industry has a welding process that can provide high quality and defect free welds in the high strength 2xxx and 7xxx alloys, development of improved alloys which can be welded without loss of properties is required, and this is a major challenge for the aluminium producers.

Although the process is asymmetrical in terms of material movement and heat generation, there is limited evidence that this has an adverse effect on the properties of the weld. Aluminium is an excellent conductor of heat, so thermal gradients will rapidly disappear. The significant heat transport in the rapidly moving material will again help minimise differences. Although many welds will have slight differences in properties on each side, this is seldom significant, and probably influenced as much by the thermodynamics and kinetics of the alloys as by the asymmetry of the process.

In a similar vein, there was considerable interest in the early days of FSW in the so called 'onion rings' in the weld. Several suggestions have been made regarding the origin of these, and evidence has been found of changes in grain size, texture, precipitation density, composition, etc., the exact mechanism varying from alloy to alloy. To date, little evidence has emerged to suggest they are harmful, and indeed their impact on weld performance seems to be very small.

The present review has demonstrated the extensive research effort that continues to progress the understanding of FSW of aluminium alloys and its influence on their microstructure and properties. It identifies a number of areas that are worthwhile for further study. From an engineering perspective, there is a need to investigate the occurrence and significance of flaws in friction stir welds. In particular, the influence of tool design on flaw occurrence and the development of nondestructive testing techniques to identify flaws in both lap and butt welds would be beneficial. Metal flow modelling may have a role to play here, though capturing this aspect of the thermomechanical behaviour remains a significant challenge. Furthermore, the development of NDT techniques capable of mapping and locating the defects that arise in FSW is an area requiring further work if they are to be utilised in safety critical applications.

The review has attempted to outline the process-microstructure-property relationships that need to be considered when friction stir welding aluminium alloys. While it is clear that FSW of aluminium alloys has reached a level of maturity that enables commercial exploitation of the technique, much remains to be done before it is possible intelligently to optimise the process to tailor the microstructure and residual stress for particular service environments, whether to provide improved tensile, fatigue, creep or corrosion resistance. Development of greater understanding of friction stir processing on sheet forming capabilities would also be of interest.

Modelling of heat generation and thermal history is reasonably mature, particularly if the heat input is measured independently from the machine itself. Microstructure evolution in both heat-treatable and non-heat-treatable alloys has been modelled at various levels of complexity, enabling prediction of hardness profiles after a degree of calibration to specific alloys and tempers. Some progress has been made to package these models for industrial use by the non-expert, and for predicting more difficult properties such as toughness. Future microstructure modelling challenges include the ability to consider FS weldability in alloy development programmes, and improved understanding of dissimilar alloy FSW, particularly for joining aluminium to other alloys such as steel for automotive applications.

Comparisons with other processes are of course inevitable. The main competitors to FSW are MIG and TIG welding, and laser welding. In terms of heat input, FSW typically introduces more heat per unit length than laser welding, but less than MIG or TIG. The fact that FSW is basically a machine tool process, closely related to milling, gives it an advantage of full automation, but at the cost of the flexibility which can be achieved with MIG and TIG. The absence of a filler in FSW is an advantage in many cases, but can also be a disadvantage since it prevents the process being used for fillet welds. The absence of filler also requires a higher standard of preparation in terms of weld gap and fit-up than fusion processes. Friction stir welding has a much greater range of thickness than any fusion process (with the exception of electron beam welding). The process has been used to make single pass welds in thicknesses from about 0⋅ 5 to 100 mm. Static mechanical properties in aluminium alloys are generally related to the heat input, and so FSW generally equals or exceeds the performance of MIG welds. Dynamic properties are generally better than fusion welds, irrespective of whether the weld surfaces are dressed. Finally, as FSW is a solid state process, it can cope with any aluminium alloy, whereas many high strength aluminium alloys are challenging or impossible to weld by fusion processes. Friction stir welding therefore joins the armoury of welding processes, but it will not displace other established processes in all applications. It has, however, succeeded in making the welding of high strength alloys a reality.

In the future, it seems likely that FSW will continue to displace MIG in applications requiring long, essentially straight welds and is well suited to the joining of aluminium extrusions and panels. It is expected to encroach more gradually on MIG for lower volume and more complex joints. However, MIG is likely to be retained in cases requiring a filler or where a manual/portable process is preferred. The barriers to growth of FSW for Al are primarily high cost of bespoke equipment (almost all machines built so far have been one-off designs (e.g. Fig.2a), but low cost milling machines can be modified), under utilisation of expensive equipment, licence fees, lack of familiarity with the process by customers, plus poor availability of skilled practitioners with industrial experience, regulatory authorities, etc. All these issues are gradually being eroded, making FSW increasingly attractive. In addition FSW is becoming indispensible where it is necessary to join different aluminium alloys. Finally, it is worth pointing to the steeply increasing volume of work looking at FSW of a wide range of systems beyond aluminium alloys. This suggests that commercial applications of FSW are set to expand widely in coming years.

Acknowledgements

The authors are deeply indebted to many colleagues within the global FSW community for numerous direct and indirect contributions to this work, for making available micrographs and figures and for many useful discussions over the past few years. PLT and AJL were afforded time by TWI to write this review. PJW is grateful to Dr Altenkirch and Dr Steuwer for assistance in providing some of the data presented.

References

1. W. M. Thomas, E. D. Nicholas, J. C. Needham, M. G. Murch, P. Temple-Smith and C. J. Dawes: 'Friction stir butt welding', GB patent no. 9125978⋅ 8, 1991.
2. W. M. Thomas, E. D. Nicholas, J. C. Needham, M. G. Murch, P. Temple-Smith and C. J. Dawes: 'Improvements relating to friction welding', US patent no. 5 460 317; EPS 0 616 490, 1991.
3. I. J. Smith and D. D. R. Lord: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI, Paper no. 2007-01-1707.
4. P. L. Threadgill: Sci. Technol. Weld. Join., 2007, 12, 357-360.
5. W. Tang, X. Guo, J. C. McClure and L. E. Murr: J. Mater. Process. Manuf. Sci., 1999, 7, 163-172.
6. M. W. Mahoney, C. G. Rhodes, J. G. Flintoff, R. A. Spurling and W. H. Bingel: Metall. Mater. Trans. A, 1998, 29A, 1955-1964.
7. A. P. Reynolds, W. D. Lockwood and T. U. Seidel: Mater. Sci. Forum, 2000, 331-337, 1719-1724.
8. L. E. Murr, G. Liu and J. C. McClure: J. Mater. Sci., 1998, 33, 1243-1251.
9. G. J. Bendszak, T. H. North and C. B. Smith: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
10. K. A. A. Hassan, P. B. Prangnell, A. F. Norman, D. A. Price and S. W. Williams: Sci. Technol. Weld. Join., 2003, 8, 257-268.
11. Y. K. Yang, H. Dong and S. Kou: Weld. J., 2008, 87, 202s-211s
12. A. Gerlich, M. Yamamoto and T. H. North: Sci. Technol. Weld. Join., 2007, 12, 472-480.
13. J. H. Yan, M. A. Sutton and A. P. Reynolds: Sci. Technol. Weld. Join., 2005, 10, 725-736.
14. P. A. Colegrove, H. R. Shercliff and R. Zettler: Sci. Technol. Weld. Join., 2007, 12, 284-297.
15. C.-G. Andersson, R. E. Andrews, B. G. I. Dance, M. J. Russell, E. J. Olden and R. M. Sanderson: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
16. L. Cedeqvist and R. E. Andrews: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI, 1400-1430.
17. K. Savolainen, J. Mononen, T. Saukkonen, H. Hänninen and J. Koivula: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI, 19.
18. W. B. Lee and S. B. Jung: Mater. Lett., 2004, 58, 1041-1046.
19. H. S. Park, T. Kimura, T. Murakami, Y. Nagano, K. Nakata and M. Ushio: Mater. Sci. Eng. A, 2004, A371, 160-169.
20. T. Sakthivel and J. Mukhopadhyay: J. Mater. Sci., 2007, 42, 8126-8129.
21. G. M. Xie, Z. Y. Ma and L. Geng: Scr. Mater., 2007, 57, 73-76.
22. P. Volovitch, J.-E. Masse, T. Baudin, B. da Costa, J. C. Goussain, W. Saikaly and L. Barrallier: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
23. J. A. Esparza, W. C. Davis, E. A. Trillo and L. E. Murr: Mater. Sci. Lett., 2002, 21, 917-920.
24. R. Johnson: Mater. Sci. Forum, 2003, 419-422, 365-370.
25. P. L. Threadgill and R. Johnson: Proc. Symp. on 'Magnesium technology', San Diego, CA, USA, March 2003, TMS.
26. W. B. Lee, J. W. Kim, Y. M. Yeon and S. B. Jung: Mater. Trans., 2003, 44, 917-923.
27. S. H. C. Park, Y. S. Sato and H. Kokawa: Scr. Mater., 2003, 49, 161-166.
28. S. H. C. Park, Y. S. Sato, H. Kokawa and T. Tsukeda: in 'Trends in welding research', (ed. S. A. David et al.), 267-272; 2003, Materials Park, OH, ASM International.
29. N. Afrin, D. L. Chen, X. Cao and M. Jahazi: Mater. Sci. Eng. A, 2008, A472, 179-186.
30. W. B. Lee, C. Y. Lee, W. S. Chang, Y. M. Yeon and S. B. Jung: Mater. Lett., 2005, 59, 3315-3318.
31. A. J. Ramirez and M. C. Juhas: Mater. Sci. Forum, 2003, 426-432, 2999-3004.
32. A. P. Reynolds, E. Hood and W. Tang: Scr. Mater., 2005, 52, 491-494.
33. B. P. Wynne, P. S. Davies, M. J. Thomas, B. S. Ng and P. L. Threadgill: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
34. M. J. Russell, P. L. Threadgill, M. J. Thomas and B. P. Wynne: Proc. 11th Int. Conf. on 'Titanium', Kyoto, Japan, June 2007, Japan Institute of Metals, 1095-1098.
35. W. M. Thomas, P. L. Threadgill and E. D. Nicholas: Sci. Technol. Weld. Join., 1999, 4, 365-372.
36. R. Johnson and P. L. Threadgill: Proc. 6th Int. Conf. on 'Trends in welding research', Pine Mountain, GA, USA, June 2002, TWI.
37. T. J. Lienert, W. L. Stellwag, B. B. Grimmett and R. W. Warke: Weld. J., 2003, 82, 1S-9S.
38. S. H. C. Park, Y. S. Sato, H. Kokawa, K. Okamoto, S. Hirano and M. Inagaki: Scr. Mater., 2003, 49, 1175-1180.
39. A. P. Reynolds, W. Tang, T. Gnaupel-Herold and H. Prask: Scr. Mater., 2003, 48, 1289-1294.
40. A. P. Reynolds, W. Tang, M. Posada and J. DeLoach: Sci. Technol. Weld. Join., 2003, 8, 455-460.
41. H. Fujii, L. Cui, N. Tsuji, M. Maeda, K. Nakata and K. Nogi: Mater. Sci. Eng. A, 2006, A429, 50-57.
42. L. Cui, H. Fujii, N. Tsuji and K. Nogi: Scr. Mater., 2007, 56, 637-640.
43. S. J. Barnes, A. Steuwer, S. Mahawish, R. Johnson and P. J. Withers: Mater. Sci. Eng. A, 2008, A492, 35-44.
44. H. J. Jun and R. Ayer: Proc. 6th Int. Symp. on 'Friction stir welding', Saint-Sauveur, Montreal, Canada, October 2006, TWI.
45. H. J. Jun, R. Ayer, T. Neeraj and R. Steel: Mater. Sci. Forum, 2007, 539-543, 3763-3768.
46. F. X. Ye, H. Fujii, T. Tsumura and K. Nakata: J. Mater. Sci., 2006, 41, 5376-5379.
47. H. Fujii, H. Kato, K. Nakata and K. Nogi: Proc. 6th Int. Symp. on 'Friction stir welding', Saint-Sauveur, Montreal, Canada, October 2006, TWI.
48. G. Biallas, R. Braun, C. Dalle Donne, G. Staniek and W. A. Kaysser: Proc. Conf. 1st Int. Symp. on ;Friction stir welding;, Thousand Oaks, CA, USA, June 1999, TWI.
49. H. Larsson, L. Karlsson, S. Stoltz and E.-L. Bergqvist: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
50. S. Tanaka and M. Kumagai: Proc. 3rd Int. Symp. on 'Friction stir welding', Kobe, Japan, September 2001, TWI.
51. H. R. Shercliff, M. J. Russell, A. D. Taylor and T. L. Dickerson: Mécan. Indust., 2005, 6, 25-35.
52. M. J. Peel, A. Steuwer and P. J. Withers: Metall. Mater. Trans. A, 2006, 37A, 2195-2206.
53. M. J. Peel, A. Steuwer, P. J. Withers, T. Dickerson, Q. Shi and H. Shercliff: Metall. Mater. Trans. A, 2006, 37A, 2183-2193.
54. A. Steuwer, M. J. Peel and P. J. Withers: Mater. Sci. Eng. A, 2006, A441, 187-196.
55. F. Palm: Proc. Conf. Materials Week '99, Munich, Germany, September 1999, DGM.
56. W. B. Lee, Y. M. Yeon and S. B. Jung: Scr. Mater., 2003, 49, 423-428.
57. Y. Li, L. E. Murr and J. C. McClure: Mater. Sci. Eng. A, 1999, A271, 213-223.
58. Y. Li, L. E. Murr and J. C. McClure: Scr. Mater., 1999, 40, 1041-1046.
59. S. Lim, S. Kim and C. G. Lee: Metall. Mater. Trans. A, 2004, 35A, 2837-2843.
60. O. T. Midling: Proc. 4th Int. Conf. on 'Aluminium alloys', Atlanta, GA, USA, September 1994, Georgia Institute of Technology, 451-458.
61. J. H. Ouyang and R. Kovacevic: J. Mater. Eng. Perform., 2002, 11, 51-63.
62. A. Gerlich, P. Su and T. H. North: Sci. Technol. Weld. Join., 2005, 10, 647-652.
63. S. A. Khodir and T. Shibayanagi: Mater. Trans., 2007, 48, 2501-2505.
64. Y. S. Sato, S. H. C. Park, M. Michiuchi and H. Kokawa: Scr. Mater., 2004, 50, 1233-1236.
65. A. C. Somasekharan and L. E. Murr: Mater. Charact., 2004, 52, 49-64.
66. J. A. Wert: Scr. Mater., 2003, 49, 607-612.
67. C. M. Chen and R. Kovacevic: Int. J. Mach. Tools Manuf., 2004, 44, 1205-1214.
68. K. Kimapong and T. Watanabe: Weld. J., 2004, 83, 277S-282S.
69. H. Uzun, C. D. Donne, A. Argagnotto, T. Ghidini and C. Gambaro: Mater. Design, 2005, 26, 41-46.
70. W. B. Lee and S. B. Jung: Mater. Res. Innov., 2004, 8, 93-96.
71. L. E. Murr, Y. Li, R. D. Flores, E. A. Trillo and J. C. McClure: Mater. Res. Innov., 1998, 2, 150-163.
72. L. E. Murr, R. D. Flores, O. V. Flores, J. C. McClure, G. Liu and D. Brown: Mater. Res. Innov., 1998, 1, 211-223.
73. L. E. Murr, Y. Li, E. A. Trillo, R. D. Flores and J. C. McClure: J. Mater. Process. Manuf. Sci., 1999, 7, 145-161.
74. R. S. Mishra and Z. Y. Ma: Mater. Sci. Eng. R, 2005, R50, 1-78.
75. R. Nandan, T. DebRoy and H. K. D. H. Bhadeshia: Prog. Mater. Sci., 2008, 53, 980-1023.
76. R. S. Mishra and M. W. Mahoney (eds.): in 'ASM specialty handbook: friction stir welding and processing'; 2007, Materials Park, OH, ASM International.
77. H. R. Shercliff and P. A. Colegrove: in 'Mathematical modelling of weld phenomena 6', (ed. H. Cerjak), 927-974; 2002, London, The Institute of Metals.
78. H. R. Shercliff and P. A. Colegrove: in 'Friction stir welding and processing', (ed. R. S. Mishra et al.), 187-217; 2007, Materials Park, OH, ASM International.
79. D. A. Price, S. W. Williams, A. Wescott, C. J. C. Harrison, A. Rezai, A. Steuwer, M. Peel, P. Staron and M. Kocak: Sci. Technol. Weld. Join., 2007, 12, 620-633.
80. M. M. Z. Ahmed, B. P. Wynne, W. M. Rainforth and P. L. Threadgill: Scr. Mater., 2008, 59, 507-510.
81. P. L. Threadgill and M. E. Nunn: 'A review of friction stir welding: part 1 process overview ', TWI members report no. 760/2003, TWI, Abington, UK, 2003.
82. Y. Uematsu, K. Tokaji, Y. Tozaki, T. Kurita and S. Murata: Int. J. Fatig., 2008, 30, 1956-1966.
83. O. T. Midling, J. S. Kvåle and O. Dahl: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
84. D. Gesto, V. Pintos, J. Vazquez, J. Rasilla and S. Barreras: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
85. 'Super high-speed passenger cargo TSL successfully completed its sea trial', www.mes.co.jp/english/ 2005 (accessed September 2008).
86. J. Ray: 'Delta 4 fleet goes from: 'Medium' to 'Heavy'', in 'Spaceflight Now', 2002.
87. M. R. Johnsen: Weld. J., 1999, 78, 35-39.
88. D. J. Waldron and R. W. Roberts: Proc. Conf. on 'Aerospace automated fastening', Long Beach, CA, USA, September 1998, SAE, 15-17.
89. J. Ding, R. Carter, K. Lawless, A. Nunes, C. Russell, M. Suites and J. Schneider. 'A decade of friction stir welding R and D at NASA's Marshall Space Flight Center and a glance into the future', ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20080009619_2008009118.pdf 2005 (accessed September 2008).
90. C. Dawes: Proc. AIAA Int. Air and Space Symp., Dayton, OH, USA, July 2003, AIAA, AIAA-2003-2769.
91. Z. S. Loftus, W. J. Arbegast and P. J. Hartley: Proc. 5th Int. Conf. on 'Trends in welding research', Pine Mountain, GA, USA, June 1998, ASM International, 580.
92. B. Christner, J. McCoury and S. Higgins: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
93. T. Kawasaki, T. Makino, S. Todori, H. Takai, M. Ezumi and Y. Ina: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
94. D. Otsuka and Y. Sakai: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
95. M. M. Shahri and R. Sandström: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
96. Anon: Mach. Design, 2003, 75, S2.
97. J. C. Bassett and S. S. Birley: Proc 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
98. G. Campbell and T. Stotler: Weld. J., 1999, 78, 45-47.
99. R. S. Mishra and M. W. Mahoney: in 'Superplasticity in advanced materials: ICSAM-2000', 507-512; 2001, Zurich-Uetikon, Trans Tech Publications Ltd.
100. J. Q. Su, T. W. Nelson and C. J. Sterling: Scr. Mater., 2005, 52, 135-140.
101. R. S. Mishra, M. W. Mahoney, S. X. McFadden, N. A. Mara and A. K. Mukherjee: Scr. Mater., 1999, 42, 163-168.
102. P. B. Berbon, W. H. Bingel, R. S. Mishra, C. C. Bampton and M. W. Mahoney: Scr. Mater., 2001, 44, 61-66.
103. J. C. Bersaas, A. Oosterkamp and L. D. Oosterkamp: 'Friction stir welding method and apparatus', Patent no. WO/2001/028732, 2001.
104. P. Su, A. Gerlich, T. H. North and G. J. Bendzsak: Sci. Technol. Weld. Join., 2006, 11, 61-71.
105. W. S. Chang, C. K. Chun, H. J. Kim, H. J. Cho and T. K. Kim: Proc. IWJC-Korea 2007, Seoul, South Korea, May 2007, Genicom Co., Ltd, 435-438.
106. M. Mahoney, R. S. Mishra, T. Nelson, J. Flintoff, R. Islamgaliev and Y. Hovansky: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 183-194; 2001, Warrendale, PA, TMS.
107. I. Charit, R. S. Mishra and M. W. Mahoney: Scr. Mater., 2002, 47, 631-636.
108. I. Charit, R. S. Mishra and K. V. Jata: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 225-234; 2001, Warrendale, PA, TMS.
109. I. Shigematsu, N. Saito, T. Komaya, T. Tamaki, G. Yamauchi and M. Nakamura: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 217-224; 2001, Warrendale, PA, TMS.
110. J. Zheng, R. S. Mishra, P. B. Berbon and M. W. Mahoney: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 235- 242; 2001, Warrendale, PA, TMS.
111. T. U. Seidel and A. P. Reynolds: Sci. Technol. Weld. Join., 2003, 8, 175-183.
112. P. A. Colegrove and H. R. Shercliff: Sci. Technol. Weld. Join., 2004, 9, 483-492.
113. A. P. Reynolds: Sci. Technol. Weld. Join., 2000, 5, 120-124.
114. W. M. Thomas, E. D. Nicholas and S. D. Smith: in 'Aluminum 2001', (ed. S. K. Das et al.), 213-224; 2001, Warrendale, PA, TMS.
115. P. A. Colegrove and H. R. Shercliff: Sci. Technol. Weld. Join., 2004, 9, 352-361.
116. D. Allehaux and F. Marie: Mater. Sci. Forum, 2006, 519-521, 1131-1138.
117. C. B. Fuller: in 'Friction stir welding and processing', (ed. R. S. Mishra et al.), 7-35; 2007, Materials Park, OH, ASM International.
118. L. Dubourg and P. Dacheux: Proc. 6th Int. Symp. on 'Friction stir welding', Saint-Sauveur, Mont., Canada, October 2006, TMS.
119. A. Bastier, M. H. Maitournam, K. D. Van and F. Roger: Sci. Technol. Weld. Join., 2006, 11, 278-288.
120. K. Colligan: Weld. J., 1999, 78, 229S-237S.
121. O. Frigaard, O. Grong and O. T. Midling: Metall. Mater. Trans. A, 2001, 32A, 1189-1200.
122. M. J. Russell, H. R. Shercliff and P. L. Threadgill: in 'Aluminum 2001', (ed. S. K. Das et al.), 2001, Warrendale, PA, TMS.
123. Y. J. Chao, X. H. Qi and W. Tang: Trans. ASME J. Manuf. Sci. Eng., 2003, 125, 138-145.
124. M. Song and R. Kovacevic: Proc. Inst. Mech. Eng. B, 2003, 217B, 73-85.
125. A. P. Reynolds, Z. Khandkar, T. Long, W. Tang and J. Khan: Mater. Sci. Forum, 2003, 426-432, 2959-2964.
126. H. Schmidt, J. Hattel and J. Wert: Model. Simul. Mater. Sci. Eng., 2004, 12, 143-157.
127. R. Kovacevic, V. Soundararajan and S. Zekovic: Int. J. Mach. Tools Manuf., 2005, 45, 1577-1587.
128. Q.-Y. Shi, T. L. Dickerson and H. R. Shercliff: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
129. A. Simar, T. Pardoen and B. Meester: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
130. P. A. Colegrove and H. R. Shercliff: Sci. Technol. Weld. Join., 2006, 11, 429-441.
131. V. I. Vill': 'Friction welding of metals'; 1959, Mashgiz, Leningrad.
132. R. E. Andrews and K. A. Beamish: 'Characterisation of and guidelines for rotary friction welding of common metallic engineering materials', TWI members report no. 824, TWI, Abington, UK, 2005.
133. Y. G. Kim, H. Fujii, T. Tsumura, T. Komazaki and K. Nakata: Mater. Sci. Eng. A, 2006, A415, 250-254.
134. P. L. Threadgill: TWI Bull., 1997, 28, 30-33.
135. P. L. Threadgill and A. J. Leonard: 'Macro and microstructural features of friction stir welds in various materials', TWI members report no. 693/1999, TWI, Abington, UK, 1999.
136. 'Specification for friction stir welding of aluminum alloys for aerospace applications', Standard D17⋅ 3:200X, American Welding Society, Miami, FL, USA, 2006.
137. A. Gerlich, P. Su, M. Yamamoto and T. H. North: Sci. Technol. Weld. Join., 2008, 13, 254-264.
138. A. D. Gingell and T. G. Gooch: 'Review of factors influencing porosity in aluminium arc welds', TWI members report no. 625/1997, TWI, Abington, UK, 1997.
139. M. F. Gittos and M. H. Scott: TWI Bull., 1987, 28, 259-263.
140. A. J. Leonard and S. A. Lockyer: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
141. T. Hashimoto, S. Jyogan, K. Nakata, Y. G. Kiu and M. Ushio: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
142. Y. S. Sato, M. Urata and H. Kokawa: Metall. Mater. Trans. A, 2002, 33A, 625-635.
143. B. K. Christner and G. D. Sylva: Proc. Conf. ICAWT '96, Columbus, OH, USA, November 1996, EWI, 359-368.
144. S. K. Chimbli, D. J. Medlin and W. J. Arbegast: in 'Friction stir welding and processing IV', (ed. R. S. Mishra et al.), 135-142; 2007, Warrendale, PA, TMS.
145. V. Balasubramanian: Mater. Sci. Eng. A, 2008, A480, 397-403.
146. K. Kumar, S. V. Kailas and T. S. Srivatsan: Mater. Manuf. Process., 2008, 23, 189-195.
147. S. T. Wei, C. Y. Hao and J. C. Chen: Mater. Sci. Eng. A, 2007, A452, 170-177.
148. Y. H. Zhao, S. B. Lin, L. Wu and F. X. Qu: Mater. Lett., 2005, 59, 2948-2952.
149. R. Crawford, G. E. Cook, A. M. Strauss, D. A. Hartman and M. A. Stremler: Sci. Technol. Weld. Join., 2006, 11, 657-665.
150. O. Lorrain, V. Favier, H. Zahrouni and M. E. Hadrouz: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, 2008, TWI.
151. H. Schmidt and J. Hattel: in 'Friction stir welding processing III', (ed. K. V. Jata et al.), 225-232; 2005, Warrendale, PA, TMS.
152. H. Schmidt and J. Hattel: Model. Simul. Mater. Sci. Eng., 2005, 13, 77-93.
153. H. Jin, S. Saimoto, M. Ball and P. L. Threadgill: Mater. Sci. Technol., 2001, 17, 1605-1614.
154. H. Okamura, K. Aota, M. Sakamoto, M. Ezumi and K. Ikeuchi: Weld. Int., 2002, 16, 266-275.
155. F. Palm, H. Steiger and U. Henneböhle: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
156. T. Jene, G. Dobmann, G. Wagner and D. Eifler: Proc. 6th Int. Symp. on 'Friction stir welding', Saint-Sauveur, Montreal, Canada, October 2006, TWI.
157. K. Savolainen, T. Saukkonen, J. Mononen and H. Hänninen: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
158. Y. S. Sato, F. Yamashita, Y. Sugiura, S. H. C. Park and H. Kokawa: Scr. Mater., 2004, 50, 365-369.
159. H. Okamura, K. Aota, M. Sakamoto, M. Ezumi and K. Ikeuchi: J. Jpn Weld. Soc., 2001, 19, 446-456.
160. T. L. Dickerson, H. R. Shercliff and H. Schmidt: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
161. C. Z. Zhou, X. Q. Yang and G. H. Luan: Scr. Mater., 2006, 54, 1515-1520.
162. J. Pryzdatek: Proc 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
163. A. J. Leonard: Unpublished work, TWI Ltd, 2001.
164. C. J. Goodfellow and A. J. Leonard: Unpublished work, TWI and Birmingham University, 2003.
165. R. Johnson: 'Further assessment of the friction stir welding of magnesium alloys', TWI members report no. 766/2003, TWI, Abington, UK, 2003.
166. L. Cederqvist and A. R. Reynolds: Weld. J., 2001, 80, 281S-287S.
167. W. M. Thomas, D. G. Staines, I. M. Norris and R. de Frias: Proc. FSW Semin., Porto, Portugal, December 2002, IST.
168. P. A. Colegrove, T. Hyoe and H. R. Shercliff: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
169. D. Lévesque, C. Mandache, L. Dubourg and P. Gougeon: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
170. C. R. Bird: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, Utah, USA, May 2003, TWI.
171. C. R. Bird: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
172. T. Vugrin, G. Staniek, W. Hillger and C. Dalle Donne: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
173. M. Moles and A. Lamarre: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
174. N. Goldfine, D. Grundy, V. Zilberstein and D. G. Kinchen: Proc. 6th Int. Conf. on 'Trends in welding research', Pine Mountain, GA, April 2002, ASM International.
175. N. Oiwa, S. Iwaki, T. Okada, N. Eguchi, S. Tanaka and K. Namba: Proc. Int. Conf. on 'Aluminium connections', Cleveland, OH, USA, June 2004, Lincoln Electric Company.
176. A. Lamarre, O. Dupuis and M. Moles: Proc. WCNDT 2004, Montreal, Canada, August-September 2004, Paper 84.
177. A. Steuwer, M. Dumont, J. Altenkirch, S. Birosca, A. Deschamps, P. B. Prangnell and P. J. Withers: 'Friction stir welding Al-Li AA2199: I microstructural aspects', In preparation, 2008.
178. A. J. Leonard: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
179. J. Karlsson, B. Karlsson, H. Larsson, L. Karlsson and L.-E. Svensson: Proc. 7th Int. Conf. on 'Joints in aluminium', Cambridge, UK, April 1998, TWI, 221-230.
180. M. Peel, A. Steuwer, M. Preuss and P. J. Withers: Acta Mater., 2003, 51, 4791-4801.
181. H. W. Hayden, S. A. David, S. S. Babu, S. Spooner, J. M. Vitek and P. J. Hartley: Prof. Conf. 3rd Int. Forum on 'Aluminium ships', Haugesund, Norway, May 1998.
182. J. Q. Su, T. W. Nelson, R. Mishra and M. Mahoney: Acta Mater., 2003, 51, 713-729.
183. A. F. Norman, I. Brough and P. B. Prangnell: Mater. Sci. Forum, 2000, 331-333, 1713-1718.
184. M. M. Attallah, C. L. Davis and M. Strangwood: Sci. Technol. Weld. Join., 2007, 12, 361-369.
185. K. N. Krishnan: Mater. Sci. Eng. A, 2002, A327, 246-251.
186. T. U. Seidel and A. P. Reynolds: Metall. Mater. Trans. A, 2001, 32A, 2879-2884.
187. M. Strangwood, J. E. Berry, D. P. Cleugh, A. J. Leonard and P. L. Threadgill: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
188. L.-E. Svensson and L. Karlsson: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
189. T. J. Lienert and R. J. Grylls: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
190. C. G. Rhodes, M. W. Mahoney, W. H. Bingel and M. Calabrese: Scr. Mater., 2003, 48, 1451-1455.
191. K. V. Jata, K. K. Sankaran and J. J. Ruschau: Metall. Mater. Trans. A, 2000, 31A, 2181-2192.
192. Y. S. Sato, H. Kokawa, M. Enomoto and S. Jogan: Metall. Mater. Trans. A, 1999, 30A, 2429-2437.
193. K. A. A. Hassan, B. P. Wynne and P. B. Prangnell: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, May 2003, TWI.
194. H. S. Yang: Proc. Conf. 6th Int. Conf. on 'Aluminium alloys', Toyohashi, Japan, July 1998, The Japan Institute of Light Metals, 1483-1488.
195. Ø. Frigaard, Ø. Grong, J. Hjelen, S. Gulbrandsen-Dahl and O. T. Midling: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
196. R. W. Fonda, J. F. Bingert and K. J. Colligan: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
197. R. S. Mishra, R. K. Islamgaliev, T. W. Nelson, Y. Hovansky and M. W. Mahoney: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 205-216; 2001, Warrendale, PA, TMS.
198. K. V. Jata and S. L. Semiatin: Scr. Mater., 2000, 43, 743-749.
199. P. B. Prangnell and C. P. Heason: Acta Mater., 2005, 53, 3179-3192.
200. Y. S. Sato, H.Watanabe, S. H. C. Park and H. Kokawa: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
201. R. S. Mishra, R. K. Islamgaliev, T. W. Nelson, Y. Hovansky and M. W. Mahoney: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 205-216; 2001, Warrendale, PA, TMS.
202. E. Litwinski: Proc. 3rd Int. Symp. on 'Friction stir welding', Kobe, Japan, September 2001, TWI.
203. M. M. Attallah and H. G. Salem Hassan: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
204. Y. S. Sato and H. Kokawa: Metall. Mater. Trans. A, 2001, 32A, 3023-3031.
205. K. N. Krishnan: J. Mater. Sci., 2002, 37, 473-480.
206. M. Karlsen, S. Tangen, J. Hjelen, Ø. Frigaard and Ø. Grong: Proc. 3rd Int. Symp. on 'Friction stir welding', Kobe, Japan, September 2001, TWI.
207. K. A. A. Hassan, A. F. Norman, D. A. Price and P. B. Prangnell: Acta Mater., 2003, 51, 1923-1936.
208. F. J. Humphreys and M. Hatherly: 'Recrystallisation and related annealing phenomena'; 1995, Oxford, Pergamon.
209. F. J. Humphreys: Acta Mater., 1997, 45, 5031-5039.
210. F. J. Humphreys: Acta Mater., 1997, 45, 4231-4240.
211. M. F. Gittos and K. Bridges: 'A study of arc and friction stir welding of two Al alloys containing a low level scandium addition', TWI members report no. 776/2003, TWI, Abington, UK, 2003.
212. B. Huneau, X. Sauvage, S. Marya and A. Poitou: in 'Friction stir welding processing III', (ed. K. V. Jata et al.), 253-260; 2005, Warrendale, PA, TMS.
213. C. S. Paglia, K. V. Jata and R. G. Buchheit: Mater. Sci. Eng. A, 2006, A424, 196-204.
214. Y. S. Sato, M. Urata, H. Kokawa and K. Ikeda: Scr. Mater., 2002, 47, 869-873.
215. I. Charit and R. S. Mishra: Acta Mater., 2005, 53, 4211-4223.
216. C. J. Hsu, C. Y. Chang, P. W. Kao, N. J. Ho and C. P. Chang: Acta Mater., 2006, 54, 5241-5249.
217. Y. S. Sato, H. Kokawa, K. Ikeda, M. Enomoto, S. Jogan and T. Hashimoto: Metall. Mater. Trans. A, 2001, 32A, 941-948.
218. D. P. Field, T. W. Nelson, Y. Hovanski and K. V. Jata: Metall. Mater. Trans. A, 2001, 32A, 2869-2877.
219. T. W. Nelson, B. Hunsaker and D. P. Field: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
220. O. V. Flores, C. Kennedy, L. E. Murr, D. Brown, S. Pappu, B. M. Nowak and J. C. McClure: Scr. Mater., 1998, 38, 703-708.
221. H. Larsson, L. Karlsson and L.-E. Svensson: Proc. 6th Int. Conf. on 'Aluminium alloys', Toyohashi, Japan, July 1998, The Japan Institute of Light Metals, 1471-1476.
222. C. Genevois, A. Deschamps and A. Denquin: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
223. M. M. Attallah, C. L. Davis and M. Strangwood: J. Mater. Sci., 2007, 42, 7299-7306.
224. R. W. Fonda and J. F. Bingert: in 'Friction stir welding and processing II', (ed. K. V. Jata et al.), 191-198; 2003, Warrendale, PA, TMS.
225. C. G. Rhodes, M. W. Mahoney, W. H. Bingel, R. A. Spurling and C. C. Bampton: Scr. Mater., 1997, 36, 69-75.
226. M. Kumagai and S. Tanaka: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
227. Y. S. Sato, H. Kokawa, M. Enomoto, S. Jogan and T. Hashimoto: Metall. Mater. Trans. A, 1999, 30A, 3125-3130.
228. Y. S. Sato, S. H. C. Park and H. Kokawa: Metall. Mater. Trans. A, 2001, 32A, 3033-3042.
229. J. D. Robson, A. Sullivan, H. R. Shercliff and G. McShane: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
230. M. Dumont, A. Steuwer, A. Deschamps, M. Peel and P. J. Withers: Acta Mater., 2006, 54, 4793-4801.
231. R. W. Fonda, J. F. Bingert and K. J. Colligan: Scr. Mater., 2004, 51, 243-248.
232. C. Genevois, A. Deschamps, A. Denquin and B. Doisneaucottignies: Acta Mater., 2005, 53, 2447-2458.
233. C. J. Dawes: TWI Bull., 2000, 41, 51-55.
234. K. Lindner, Z. Khandkar, J. Khan, W. Tang and A. P. Reynolds: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
235. R. Y. Hwang and C. P. Chou: Scr. Mater., 1997, 38, 215-221.
236. A. Sullivan and J. D. Robson: Mater. Sci. Eng. A, 2008, A478, 351-360.
237. L. E. Svensson, L. Karlsson, H. Larsson, B. Karlsson, M. Fazzini and J. Karlsson: Sci. Technol. Weld. Join., 2000, 5, 285-296.
238. O. R. Myhr and Ø. Grong: Acta Metall. Mater, 1991, 39, 2703-2711.
239. O. R. Myhr and Ø. Grong: Acta Metall. Mater, 1991, 39, 2693-2702.
240. O. R. Myhr, O. Grong, S. Klokkehaug, H. G. Fjoer and A. O. Kluken: Sci. Technol. Weld. Join., 1997, 2, 245-253.
241. Ø. Grong and H. R. Shercliff: Prog. Mater. Sci., 2002, 47, 163-282.
242. O. Grong and O. R. Myhr: Acta Mater., 2000, 48, 445-452.
243. O. R. Myhr and O. Grong: Acta Mater., 2000, 48, 1605-1615.
244. M. Nicolas and A. Deschamps: Acta Mater., 2003, 2003, 6077- 6094.
245. T. Hyoe, P. A. Colegrove and H. R. Shercliff: in 'Friction stir welding and processing II', (ed. K. V. Jata et al.), 33-42; 2003, Warrendale, PA, TMS.
246. J. D. Robson, A. Sullivan, G. McShane and H. R. Shercliff: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
247. C. Gallais, A. Denquin, A. Pic, A. Simar and T. Pardoen: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
248. P. A. Colegrove, H. R. Shercliff, J. D. Robson, N. Kamp, A. Sullivan and S. W. Williams: in 'Mathematical modelling of weld phenomena 8', (ed. H. Cerjak et al.), 2008, London, Maney Publishing.
249. A. Sullivan, N. Kamp and J. D. Robson: Mater. Sci. Forum, 2006, 519-521, 1181-1186.
250. N. Kamp, A. Sullivan, R. Tomasi and J. D. Robson: Acta Mater., 2006, 54, 2003-2014.
251. N. Kamp, A. Sullivan and J. D. Robson: Mater. Sci. Eng. A, 2007, A466, 246-255.
252. A. Deschamps and Y. Brechet: Acta Mater., 1998, 47, 293-305.
253. A. Deschamps, F. Livet and Y. Brechet: Acta Mater., 1998, 47, 281-292.
254. C. G. Derry and J. D. Robson: Mater. Sci. Eng. A, 2008, A490, 328-334.
255. M. W. Mahoney: in 'Friction stir welding and processing', (ed. R. S. Mishra et al.), 187-217; 2007, Materials Park, OH, ASM International.
256. H. J. Liu, H. Fujii, M. Maeda and K. Nogi: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
257. A. von Strombeck, G. Çam, J. F. dos Santos, V. Venzke and M. Koçak: in 'Aluminum 2001', (ed. S. K. Das et al.), 249-264; 2001, Warrendale, PA, TMS.
258. S. Brinkmann, A. von Strombeck, C. Schilling, J. F. dos Santos, D. Lohwasser and M. Koc¸ak: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
259. A. von Strombeck, J. F. dos Santos, F. Torster, P. Laureano and M. Koçak: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, USA, June 1999, TWI.
260. D. Allehaux, G. Petit, M.-H. Campagnac, G. Lapasset and A. Denquin: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
261. A. Denquin, D. Allehaux, H.-H. Campagnac and G. Lapasset: Proc. 3rd Int. Symp. on 'Friction stir welding', Kobe, Japan, September 2001, TWI.
262. G. Biallas and C. D. Donne: Materialprufung, 2000, 42, 236-239.
263. O. Hatamleh, J. Lyons and R. Forman: Int. J. Fatig., 2007, 29, 421-434.
264. M. A. Sutton, B. C. Yang, A. P. Reynolds and J. H. Yan: Mater. Sci. Eng. A, 2004, A364, 66-74.
265. M. J. Peel, M. Preuss, A. Steuwer, M. Turski and P. J. Withers: in 'Trends in welding research, proceedings', (ed. S. A. David et al.), 273-278; 2003, Materials Park, OH, ASM International.
266. J. Altenkirch, A. Steuwer, M. J. Peel, D. G. Richards and P. J. Withers: Mater. Sci. Eng. A, 2008, A488, 16-24.
267. J. K. Kristensen, C. Dalle Donne, T. Ghidini, J. Mononen, A. Norman, A. Pietras, M. J. Russell and S. Slater: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
268. H. J. Liu, H. Fujii, M. Maeda and K. Nogi: Mater. Sci. Technol., 2004, 20, 103-105.
269. H. J. Liu, H. Fujii, M. Maeda and K. Nogi: J. Mater. Process. Technol., 2003, 142, 692-696.
270. Lockheed Martin: presented at Aeromat '99, Dayton, OH, USA, June 1999, ASM.
271. K. Masubuchi: in 'Encyclopedia of materials: science and technology', (ed. K. H. J. Buschow et al.), 8121-8126; 2003, Oxford, Elsevier.
272. P. J. Withers and H. K. D. H. Bhadeshia: Mater. Sci. Technol., 2001, 17, 366-375.
273. J. Lu (ed.): in 'Handbook of measurement of residual stresses'; 1996, Lilburn, GA, Fairmont Press.
274. M. B. Prime: J. Mater. Eng. Technol., 2001, 123, 162-168.
275. P. J. Withers, M. Turski, L. Edwards, P. J. Bouchard and D. J. Buttle: Int. J. Press. Vess. Pip., 2008, 85, 118-127.
276. P. J. Withers: C. R. Phys., 2007, 8, 806-820.
277. P. J. Withers and P. J. Webster: Strain, 2001, 37, 19-31.
278. P. J. Withers, M. Preuss, P. J. Webster, D. J. Hughes and A. M. Korsunsky: Mater. Sci. Forum, 2002, 404-407, 1-10.
279. D. J. Buttle and C. Scruby: in 'Encyclopedia of materials: science and technology', (ed. K. H. J. Buschow et al.), 8173-8180; 2001, Oxford, Elsevier.
280. L. D. Oosterkamp, P. J. Webster, P. A. Browne, G. B. M. Vaughan and P. J. Withers: Proc. 5th Eur. Conf. on 'Residual stresses', Mater. Sci. Forum, 2000, 347-349, 687-693.
281. C. Dalle Donne: Proc. 3rd Int. Symp. on 'Friction stir welding', Kobe, Japan, September 2001, TWI.
282. P. J. Webster, L. D. Oosterkamp, P. A. Browne, D. J. Hughes, W. P. Kang, P. J. Withers and G. B. M. Vaughan: J. Strain Anal. Eng. Design, 2001, 36, 61-70.
283. M. N. James, D. J. Hughes, D. G. Hattingh, G. R. Bradley, G. Mills and P. J. Webster: Fatig. Fract. Eng. Mater. Struct., 2004, 27, 187-202.
284. X. L. Wang, Z. Feng, S. A. David, S. Spooner and C. R. Hubbard: Proc. 6th Int. Conf. on 'Residual stresses', Oxford, UK, July 2000, IOM.
285. P. Staron, M. Kocak and S. Williams: in 'Trends in welding research, proceedings', (ed. S. A. David et al.), 253-256; 2003, Materials Park, OH, ASM International.
286. M. B. Prime, T. Gnaupel-Herold, J. A. Baumann, R. J. Lederich, D. M. Bowden and R. J. Sebring: Acta Mater., 2006, 54, 4013- 4021.
287. M. Mahoney, C. Fuller, A. DeWald and M. R. Hill: Proc. 6th Int. Symp. on 'Friction stir welding', Saint-Sauveur, Montreal, Canada, October 2006, TWI.
288. T. Ghidini and C. Dalle Donne: Fatig. Fract. Eng. Mater. Struct., 2007, 30, 214-222.
289. D. Stefanescu, C. E. Truman, D. J. Smith and P. S. Whitehead: Experim. Mech., 2006, 46, 417-427.
290. P. Staron, M. Koçak, S. Williams and A. Wescott: Physica B, 2004, 350B, E491-E493.
291. R. A. Owen, R. V. Preston, P. J. Withers, H. R. Shercliff and P. J. Webster: Mater. Sci. Eng. A, 2003, A346, 159-167.
292. M. A. Sutton, A. P. Reynolds, D. Q. Wang and C. R. Hubbard: J. Eng. Mater. Technol., 2002, 124, 215-221.
293. Y. J. Chao and X. H. Qi: J. Mater. Process. Manuf. Sci., 1999, 7, 215-233.
294. C. M. Chen and R. Kovacevic: Int. J. Mach. Tools Manuf., 2003, 43, 1319-1326.
295. Z. Feng, X. L. Wang, S. A. David and P. S. Sklad: Sci. Technol. Weld. Join., 2007, 12, 348-356.
296. D. G. Richards, P. B. Prangnell, S. W. Williams and P. J. Withers: Mater. Sci. Eng., 2008, 489, 351-362.
297. A. Bastier, M. H. Maitournam, F. Roger and K. D. Van: J. Mater. Process. Technol., 2008, 200, 25-37.
298. T. D. Vuyst, V. Madhavan, B. Ducoeur, A. Simar, B. D. Meester and L. D'Alvise: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
299. M. T. Milan, W. W. Bose, J. R. Tarpani, A. M. S. Malafaia, C. P. O. Silva, B. C. Pellizer and L. E. Pereira: J. Mater. Eng. Perform., 2007, 16, 86-92.
300. H. Lombard, D. G. Hattingh, A. Steuwer and M. N. James: Mater. Sci. Eng. A, 2009, A501, 119-124.
301. J. Altenkirch, A. Steuwer, M. Poad and P. J. Withers: 'Residual stresses in 2199 as a function of welding conditions', to be submitted.
302. T. W. Greene and A. A. Holzbaur: Weld. J. Res. Suppl., 1946, 11, 171s-185s.
303. P. A. McGuire and J. J. Groom: 'Computational analysis and experimental evaluation for residual stresses from induction heating', RP1394-4, Battelle Memorial Institute, 1979.
304. P. Michaleris and X. Sun: Weld. J., 1997, 76, 451s-457s.
305. R. M. Dull, J. R. Dydo and J. J. Russell: Proc. 82nd Annual AWS Convention, Cleveland, OH, USA, May 2001, AWS, 95-96.
306. P. Dong, J. K. Hong and P. Rogers: Weld. J., 1998, 77, 439s-445s.
307. T. E. Barber, F. W. Brust, H. W. Mishler and M. F. Kanninen: 'Controlling residual stresses by heat sink welding', EPRI report no. NP-2159-LD, Electric Power Research Institute, Palo Alto, CA, USA, 1981.
308. E. M. van der Aa, M. J. M. Hermans, I. M. Richardson, N. M. van der Pers and R. Delhez: Mater. Sci. Forum, 2006, 524-525, 479-484.
309. J. Gabzdyl, M. Cole, S. W. Williams and D. Price: Proc. ICALEO 2001, Jacksonville, FL, USA, October 2001, Laser Institute of America, Vols. 92 and 93, 1236-1245.
310. J. T. Gabzdyl: 'Thermal welding', US patent no. 20010054639, 2001.
311. D. G. Richards, P. B. Prangnell, P. J. Withers, S. W. Williams and S. Morgan: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
312. G. Luan, G. Li, C. Li and C. Dong: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
313. Y. P. Yang, P. Dong, X. Tian and Z. Zhang: Proc. 5th Int. Conf. on 'Trends in welding research', (ed. S. A. David et al.), 700-705; 1998, Materials Park, OH, ASM International.
314. Y. P. Yang, P. Dong, J. Zhang and X. Tian: Weld. J. Res. Suppl., 2000, 79, 9s-17s.
315. S. W. Williams, D. A. Price, A. Wescott, C. J. C. Harrison, P. Staron and M. Kocak: Proc. Conf. on 'Welding and brazing of aerospace structures - modern applications and materials for new and in-service', Berlin, Germany, May 2004, DVS Berichte, 95-101.
316. A. L. Lafly, C. D. Donne, G. Biallas, D. Allehaux and F. Marie: Mater. Sci. Forum, 2006, 519-521, 1089-1094.
317. P. Prevey and M. Mahoney: Mater. Sci. Forum, 2003, 426-432, 2933-2939.
318. P. S. Prevey, D. J. Hornbach and N. Jayaraman: Mater. Sci. Forum, 2007, 539-547, 3807-3813.
319. A. Ali, X. An, C. A. Rodopoulos, M. W. Brown, P. O'Hara, A. Levers and S. Gardiner: Int.l J. Fatig., 2007, 29, 1531-1545.
320. O. Hatamleh, J. Lyons and R. Forman: Fatig. Fract. Eng. Mater. Struct., 2007, 30, 115-130.
321. O. Hatamleh, I. V. Rivero and J. Lyons: J. Mater. Eng. Perform., 2007, 16, 549-553.
322. N. Jayaraman, P. Prevey and M. Mahoney: in 'Friction stir welding and processing II', (ed. K. V. Jata et al.), 259-269; 2003, Warrendale, PA, TMS.
323. S. W. Wen, S. W. Williams, S. A. Morgan, A. Wescott and M. Poad: submitted to Sci. Technol. Weld. Join.
324. J. Altenkirch, A. Steuwer, P. J. Withers, S. W. Williams, M. Poad and S. Wen: Sci. Technol. Weld. Join., 2009, 14, 185-192.
325. S. W. Williams, D. A. Price, W. Wescott, A. Steuwer, M. Peel, J. Altenkirch, P. J. Withers and M. Poad: Proc. 6th Int. Symp. on 'Friction stir welding', Saint-Sauveur, Montreal, Canada, October 2006, TWI.
326. J. Altenkirch, A. Steuwer, M. J. Peel, P. J. Withers, S. Williams and M. Poad: Metall. Mater. Trans. A, 2008, 39A, 3246-3259.
327. J. Z. Zhang, R. Pedwell and H. Davies: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
328. G. E. Shepherd: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
329. L. Magnusson and L. Källman: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
330. H. Hori, S. Makita and H. Hino: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, June 1999, TWI.
331. G. Bussu and P. E. Irving: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, June 1999, TWI.
332. D. Lohwasser: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
333. M. Ericsson and R. Sandstrom: Int. J. Fatig., 2003, 25, 1379-1387.
334. P. Moreira, A. M. P. de Jesus, A. S. Ribeiro and P. de Castro: in 'Advances in fracture and damage mechanics VI', (ed. J. Alfaiate et al.), 209-212; 2007, Zurich, Trans Tech Publications.
335. M. Ericsson, R. Sandström and J. Hagström: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
336. S. Kim, C. G. Lee and S.-J. Kim: Mater. Sci. Eng. A, 2008, A478, 56-64.
337. H. Lombard, D. G. Hattingh, A. Steuwer and M. N. James: Eng. Fract. Mech., 2008, 75, 341-354.
338. C. Dalle Donne, G. Biallas, T. Ghidini and G. Raimbeaux: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
339. R. John and K. V. Jata: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 57-69; 2001, Warrendale, PA, TMS.
340. A. Cirello, G. Buffa, L. Fratini and S. Pasta: Proc. Inst. Mech. Eng. B, 2006, 220B, 805-811.
341. G. Bussu and P. E. Irving: Int. J. Fatig., 2003, 25, 77-88.
342. J. M. L. Tan, M. E. Fitzpatrick and L. Edwards: Eng. Fract. Mech., 2007, 74, 2030-2054.
343. S. S. Di, X. Q. Yang, G. H. Luan and B. Jian: Mater. Sci. Eng. A, 2006, A435, 389-395.
344. C. Z. Zhou, X. Q. Yang and G. H. Luan: J. Mater. Sci., 2006, 41, 2771-2777.
345. A. Oosterkamp, L. D. Oosterkamp and A. Nordeide: Weld. J., 2004, 83, 225S-231S.
346. A. Ali, M. W. Brown and C. A. Rodopoulos: Proc. 6th Int. Symp. on 'Friction stir welding', Saint-Sauveur, Montreal, Canada, October 2006, TWI.
347. H. B. Chen, K. Yan, T. Lin, S. B. Chen, C. Y. Jiang and Y. Zhao: Mater. Sci. Eng. A, 2006, A433, 64-69.
348. M. N. James, D. G. Hattingh and G. R. Bradley: Int. J. Fatig., 2003, 25, 1389-1398.
349. M. N. James, G. R. Bradley, H. Lombard and D. G. Hattingh: Fatig. Fract. Eng. Mater. Struct., 2005, 28, 245-256.
350. T. L. Dickerson and J. Przydatek: Int. J. Fatig., 2003, 25, 1399-1409.
351. C. A. Widener, B. M. Tweedy and D. A. Burford: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
352. C. Dalle Donne and G. Biallas: Proc. Eur. Conf. on 'Spacecraft structures, materials and mechanical testing', Noordwijk, The Netherlands, November-December 1999, ESA, 309-314.
353. S. G. Russell, M. Tester, E. Nichols, A. Cleaver and J. Maynor: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 93-104; 2001, Warrendale, PA, TMS.
354. M. T. Milan, W. W. B. Filho, C. O. F. T. Ruckert and J. R. Tarpani: Fatig. Fract. Eng. Mater. Struct., 2008, 31, 526-538.
355. C. Z. Zhou, X. Q. Yang and G. H. Luan: Scr. Mater., 2005, 53, 1187-1191.
356. H. Q. Qu, M. Tsujikawa, S. W. Chung, S. Oki and K. Higashi: in 'Recrystallization and grain growth III', (ed. S. J. L. Kang et al.), 793-796; 2007, Zurich, Trans Tech Publications.
357. M. Ranes, A. O. Kluken and O. T. Midling: Proc. 4th Int. Conf. on 'Trends in welding research', Gatlinburg, TN, USA, June 1995, ASM International.
358. S. Hong, S. Kim, C. G. Lee and S. J. Kim: J. Mater. Sci., 2007, 42, 9888-9893.
359. P. S. Pao, R. W. Fonda, H. N. Jones, C. R. Feng, B. J. Connolly and A. J. Davenport: in 'Friction stir welding processing III', (ed. K. V. Jata et al.), 27-34; 2005, Warrendale, PA, TMS.
360. P. Cavaliere, G. Campanile, F. Panella and A. Squillace: J. Mater. Process. Technol., 2006, 180, 263-270.
361. S. J. Hong, S. S. Kim, C. G. Lee and S. J. Kim: in 'Advances in nanomaterials and processing', (ed. B. T. Ahn et al.), 1321-1324; 2007, Zurich, Trans Tech Publications.
362. M. Ericsson and R. Sandstrom: Steel Res. Int., 2006, 77, 450-455.
363. P. J. Haagensen, O. T. Midling and M. Ranes: in 'Computer methods and experimental measurements for surface treatment effects II', (ed. M. H. Aliabadi et al.), 225-237; 1995, Southampton, Wit Pr/Computational Mechanics.
364. M. Kumagai, S. Tanaka, H. Hatta, H. Yoshida and H. Sato: Proc. 3rd Int. Symp. on 'Friction stir welding', Kobe, Japan, September 2001, TWI.
365. M. Ericsson and R. Sandström: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
366. W. M. Thomas, M. F. Gittos, P. Tubby, D. G. Staines and S. Lockyer: 'Friction skew stir lap welding of 5083-H111- preliminary fatigue studies', TWI members report, TWI, Abington, UK, to be published.
367. R. Pedwell, H. Davies and A. Jefferson: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, June 1999, TWI.
368. R. Talwar, D. Bolser, R. Lederich and J. Baumann: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
369. G. E. Shepherd: Proc. 4th Int. Symp. on 'Friction stir welding', Park City, UT, USA, May 2003, TWI.
370. W. M. Thomas, I. M. Norris, D. G. Staines and E. R. Watts: Proc. SME Summit, Oconomowoc,WI, USA, August 2005, SME.
371. D. Fersini and A. Pirondi: Eng. Fract. Mech., 2008, 75, 790-803.
372. D. Hornbach, M. Mahoney, P. Prevey, D. Waldron and J. Cammett: in 'Trends in welding research, proceedings', (ed. S. A. David et al.), 302-306; 2003, Materials Park, OH, ASM International.
373. S. Lomolino, R. Tovo and J. dos Santos: Int. J. Fatig., 2005, 27, 305-316.
374. M. G. Dawes, T. L. Dickerson, S. A. Karger and J. Przydatek: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
375. M. K. Kulekci, F. Mendi, I. Sevim and O. Basturk: Metalurgija, 2005, 44, 209-213.
376. M. A. Sutton, A. P. Reynolds, B. C. Yang and R. Taylor: Eng. Fract. Mech., 2003, 70, 2215-2234.
377. W. J. Arbegast, K. S. Baker and P. J. Hartley: Proc. 5th Int. Conf. on 'Trends in welding research', Pine Mountain, GA, USA, June 1998, ASM International.
378. H. R. Kroninger and A. P. Reynolds: Fatig. Fract. Eng. Mater. Struct., 2002, 25, 283-290.
379. M. Mochizuki, M. Inuzuka, H. Nishida, K. Nakata and M. Toyoda: Sci. Technol. Weld. Join., 2006, 11, 366-370.
380. C. G. Derry and J. D. Robson: Proc. 7th Int. Symp. on 'Friction stir welding', Awaji Island, Japan, May 2008, TWI.
381. M. G. Dawes, S. A. Karger and J. Przydatek: 'Fracture toughness of friction stir welds in 2014A, 7075 and 5083 aluminium alloys', TWI report no. 705/2000, TWI, Abington, UK, 2000.
382. D. G. Kinchen, Z. Li and G. P. Adams: Proc. 1st Int. Symp. on 'Friction stir welding', Thousand Oaks, CA, June 1999, TWI.
383. D. Dumont, A. Deschamps and Y. Brechet: Acta Mater., 2004, 52, 2529-2540.
384. J. Lumsden: in 'Friction Stir Welding and Processing', (ed. R. S. Mishra et al.), 187-217; 2007, Materials Park, OH, ASM International.
385. F. Hannour, A. J. Davenport and M. Strangwood: Proc. 2nd Int. Symp. on 'Friction stir welding', Gothenburg, Sweden, June 2000, TWI.
386. J. Corral, E. A. Trillo, Y. Li and L. E. Murr: J. Mater. Sci. Lett., 2000, 19, 2117-2122.
387. D. P. P. Booth, M. J. Starink and I. Sinclair: Mater. Sci. Technol., 2007, 23, 276-284.
388. A. J. Davenport, M. Jariyaboon, C. Padovani, N. Tareelap, B. J. Connolly, S. Williams and E. Siggs: Mater. Sci. Forum, 2006, 699, 519-521.
389. C. Dalle Donne, R. Braun, G. Staniek, A. Jung and W. A. Kaysser: Mater. Werkst., 1998, 29, 609-617.
390. D. B. Mitton, A. Squillace, A. De Fenzo, C. Padovani, T. Monetta, G. Giorleo, P. Cozzolino and F. Bellucci: Corros. Rev., 2007, 25, 449-459.
391. J. B. Lumsden, M. W. Mahoney, G. Pollock and C. G. Rhodes: Corrosion, 1999, 55, 1127-1135.
392. R. G. Buchheit and C. S. Paglia: in 'Corrosion and protection of light metal alloys', (ed. R. G. Buchheit et al.), 94-103; 2004, Pennington, NJ, Electrochemical Society.
393. C. S. Paglia, K. V. Jata and R. G. Buchheit: Mater. Corros., 2007, 58, 737-750.
394. X. Zhou, Y. Younes, D. Wadeson, T. Hashimoto and G. E. Thompson: Adv. Mater. Res., 2008 38, 298-305.
395. B. J. Connolly, A. J. Davenport, M. Jariyaboon, C. Padovani, R. Ambat, S. W. Williams, D. A. Price, A. Wescott, C. J. Goodfellow and C.-M. Lee: Proc. 5th Int. Symp. on 'Friction stir welding', Metz, France, September 2004, TWI.
396. M. Jariyaboon, A. J. Davenport, R. Ambat, B. J. Connolly, S. W. Williams and D. A. Price: Corros. Sci., 2007, 49, 877-909.
397. A. Squillace, A. de Fenzo, G. Giorleo and F. Bellucci: J. Mater. Process. Technol., 2004, 152, 97-105.
398. W. Hu and E. I. Meletis: Mater. Sci. Forum, 2000, 331-337, 1683-1688.
399. C. S. Paglia and R. G. Buchheit: Mater. Sci. Eng. A, 2006, A429, 107-114.
400. F. Zucchi, G. Trabanelli and V. Grassi: Mater. Corros., 2001, 52, 853-859.
401. S. Maggiolino and C. Schmid: J. Mater. Process. Technol., 2008, 197, 237-240.
402. R. Braun, C. D. Donne and G. Staniek: Mater. Werkst., 2000, 31, 1017-1026.
403. J. B. Lumsden, M. W. Mahoney, C. G. Rhodes and G. A. Pollock: Corrosion, 2003, 59, 212-219.
404. C. S. Paglia, L. M. Ungaro, B. C. Pitts, M. C. Carroll, A. P. Reynolds and R. G. Buchheit: in 'Friction stir welding and processing II', (ed. K. V. Jata et al.), 65-75; 2003, Warrendale, PA, TMS.
405. K. K. Sankaran, H. L. Smith and K. Jata: in 'Trends in welding research, proceedings', (ed. S. A. David et al.), 284-286; 2003, Materials Park, OH, ASM International.
406. D. P. Field, T. W. Nelson, Y. Hovanski and D. F. Bahr: in 'Friction stir welding and processing', (ed. K. V. Jata et al.), 83-91; 2001, Warrendale, PA, TMS.
407. C. S. Paglia, M. C. Carroll, B. C. Pitts, T. Reynolds and R. G. Buchheit: Mater. Sci. Forum, 2002, 396-402, 1677-1684.
408. X. Y. Yun, Y. Motohashi, T. Ito, T. Asano and S. Hirano: J. Jpn Inst. Met., 2006, 70, 96-105.
409. D. A. Wadeson, X. Zhou, G. E. Thompson, P. Skeldon, L. D. Oosterkamp and G. Scamans: Corros. Sci., 2006, 48, 887-897.
410. L. A. Willey: in 'Aluminium: properties, physical metallurgy and phase diagrams', (ed. K. R. van Horn), 140-158; 1967, Cleveland, OH, ASM.
411. J. R. Galvele and S. M. de Micheli: Corros. Sci., 1970, 10, 795-807.
412. J. T. Staley, S. C. Byrne, E. L. Colvin and K. P. Palmer: Mater. Sci. Forum, 1996, 217, 1587-1592.
413. C M. Lee: Corrosion, submitted.
414. C. S. Paglia and R. G. Buchheit: Scr. Mater., 2008, 58, 383-387.
415. R. Cook, T. Handboy, S. L. Fox and W. Arbegast: in 'Friction stir welding processing III', (ed. K. V. Jata et al.), 35-42; 2005, Warrendale, PA, TMS.
416. P. B. Srinivasan, W. Dietzel, R. Zettler, J. F. dos Santos and V. Sivan: Mater. Sci. Eng. A, 2005, A392, 292-300.
417. C. A. Widener, J. E. Talia, B. M. Tweedy and D. A. Burford: in 'Friction stir welding and processing IV', (ed. R. S. Mishra et al.), 449-458; 2007, Warrendale, PA, TMS.
418. M. Jariyaboon, A. J. Davenport, R. Ambat, B. J. Connolly, S. W. Williams and D. A. Price: Corros. Eng. Sci. Technol., 2006, 41, 135-142.
419. J. Lumsden, G. Pollock and M. Mahoney: Mater. Sci. Forum, 2003, 426-432, 2867-2872.
420. C. A. Widener, D. A. Burford, B. Kumar, J. E. Talia and B. Tweedy: Mater. Sci. Forum, 2007, 539-543, 3781-3788.
421. A. Merati, K. Sarda, D. Raizenne and C. D. Donne: in 'Friction stir welding and processing II', (ed. K. V. Jata et al.), 77-90; 2003, Warrendale, PA, TMS.
422. C. A. Widener, J. E. Talia, B. M. Tweedy and D. A. Burford: in 'Friction stir welding and processing IV', (ed. R. S. Mishra et al.), 459-468; 2007, Warrendale, PA, TMS.
423. J. Lumsden, G. Pollock andM.Mahoney: in 'Friction stir welding processing III', (ed. K. V. Jata et al.), 19-25; 2005, Warrendale, PA, TMS.
424. C. Padovani, L. Fratini, A. Squillace and F. Bellucci: Corros. Rev., 2007, 25, 475-489.
425. P. B. Srinivasan, W. Dietzel, R. Zettler, J. F. dos Santos and V. Sivan: Corros. Eng. Sci. Technol., 2007, 42, 161-167.
426. K. P. Rao, G. D. J. Ram and B. E. Stucker: Scr. Mater., 2008, 58, 998-1001.
427. P. Prevey, D. Hornbach, P. Mason and M. Mahoney: in 'Surface engineering: coating and heat treatments, proceedings', (ed. O. Popoola et al.), 131-137; 2003, Materials Park, OH, ASM International.