Friction stir welds in aluminium alloys - preliminary microstructural assessment

TWI Bulletin, Mar/Apr 1997

Philip Threadgill gained BSc and PhD degrees in physical metallurgy from University College, Swansea, and has worked at TWI for over twenty years.

Friction stir welding is making a huge impact on welding aluminium alloys, but as yet the microstructural features of these welds have received little attention. Philip Threadgill summarises what is known.

Introduction

Friction stir welding is arguably the most significant development in solid phase welding in the last decade. The process operates by passing a rotating tool along a joint between two closely butted sheets of material, as shown in Fig.1.

Frictional heat is developed by contact between the tool and workpieces which causes the material to soften. The motion of the tool forces the material from the front to the back of the tool, where it is consolidated to make a joint of very high quality. Details of the process have been described in recent patents and articles,[1-3] so will not be repeated here.

Most work published to date has concentrated on development and optimisation of the process, together with some information on the mechanical properties which can be obtained. This paper will start to explore some of the metallurgical features of friction stir welds in aluminium alloys, although it must be pointed out that this work is still very much in its embryonic stage. Nevertheless, sufficient information is available to allow some appreciation of the metallurgical process to be gained, and to apply this information in order to develop a more complete understanding of both the mechanics of the process, and the properties of the welds.

A number of alloys have been investigated at TWI as part of the Core Research Programme, and Group Sponsored Project 5651. Many of these welds have been the subject of microstructural examinations, which are reviewed in this article, together with work from other sources.

Macrostructure

a) Alloy 7075-T7351, AH1868;

b) Alloy 2014A-T6, AH1870

c) Alloy 5083-H321, AH1871.

The most obvious feature is the well-developed nugget which lies at the centre of the weld. This often contains a well-defined onion ring type inner structure, consisting of concentric ovals, as can be seen in these examples, although in other alloys, this feature is either not visible, or only very faintly visible. The overall shape of the nugget is very variable, depending on the alloy used and the precise process conditions. Appendages to the nugget, often of complex shape, are frequently seen at the upper surface of the weld, and can extend to the edges of the tool shoulder. The diameter of the nugget is typically slightly greater than that of the probe on the tool, and significantly less than the shoulder diameter. The nugget also typically extends to the bottom of the weld, but this does not always happen.

From current knowledge, prediction of the nugget shape is not possible. It is clearly dependent on tool design and welding parameters, and also on the hot strength of the material being welded.

The macrostructure outside the nugget is also quite distinctive. There is invariably an area adjacent to the nugget where severe plastic deformation of the material has occurred, such that the elongated grain structure of the material can be rotated by as much as 90 degrees. It has been possible to correlate these areas with the hardness levels observed in the weld, and this will be discussed along with the microstructures. Typical hardness values are shown in Fig.3.

a) Alloys 7075-T7351 and 2219-T87

b) Alloy 2014A-T6, as-welded and post-weld aged

c) Alloys 5083-O and 5083-H321

Microstructure

Examination of many friction stir welds in aluminium alloys has demonstrated that there are four major microstructural zones, as indicated in Fig.4. The outer region, Zone A in Figure 4 is parent material which is unaffected by either heat or mechanical deformation. Zone B remains undeformed, but heat from the weld has influenced the properties of the region. This is referred to as the thermally affected zone. Closer to the weld nugget, Zone C has been subjected to plastic deformation, as well as thermal effects from the weld, and is referred to as the thermo-mechanically affected zone. The nugget, Zone D, accounts for the rest of the weld.

In the parent material, Zone A, the microstructure of all alloys examined has been an elongated grain structure which reflects the rolling operation used, fig 5.

In the thermally affected zone, optical microscopy shows no apparent difference from the parent material, except that the etching response was generally more rapid. In age hardened alloys, or mechanically hardened alloys, the hardness level decreased in this area, a clear indication that heat from the welding process has a profound influence, either by overageing, or by lowering of dislocation density, and probably by both in fully aged alloys. However, there are at present no known studies of the precipitate structure in this region. In annealed material such as 5083-O, no significant change in hardness was found (Fig.3).

In the thermomechanically affected zone, it is clear that the welding process causes bending of the elongated grains, and occasionally some local recrystallisation. The thermal cycle must also be accelerating the ageing and annealing processes described earlier, although the latter will obviously now compete with some degree of work hardening. Examination of hardness traces for age hardened alloys shows that the hardness typically goes through a minimum in this region, and so the work hardening is the dominant effect close to the weld nugget, and overageing/annealing remains the most important effect towards the thermally affected zone. It is also possible that parts of this zone may reach a sufficiently high temperature to allow some of the precipitates to enter solid state solution.

However, data on temperatures achieved at various points in the weld is scant, and the extent of precipitate dissolution is difficult to estimate. There is usually significant evidence for partial recrystallisation close to the boundary with the nugget. It is interesting to note that this boundary is generally very sharply defined, perhaps slightly more so on the side where the direction of tool rotation is the same as the direction of travel. Typical examples of the microstructure from this region are shown in Fig.6.

Microstructure of the weld nugget is clearly equiaxed, and very fine. Grain sizes vary with alloy and welding procedure, but are typically less than about ten microns.

Hardness levels reported in this region are below the parent material level in age hardened and mechanically hardened alloys, but in alloys such as 5083-O, the value is slightly higher than the parent material. It is, however, interesting to note that 5083 material in the O and H321 condition gives the same hardness value in the nugget, as can be seen from Fig.4

a) Thermo-mechanically affected zone

b) Nugget area

There are some electron microscopy studies of the nugget region. The work of Rhodes et al [4] on friction stir welded 7075-T6 has shown that the equiaxed grains in the nugget are separated by high angle boundaries, demonstrating that these are not subgrains, and that recrystallisation has occurred. These authors also noted a considerable reduction in the dislocation density in the nugget compared to other regions. Rhodes et al have reported that two possible types of precipitate occur in the nugget. The first type is intragranular precipitates which could be either Mg32(Al,Zn)49 or Mg(Zn2,AlCu). At 60-80nm in size, these are slightly larger than the 50-75nm size range reported by the same authors in the parent material.

Additional precipitates, identified as Mg(Zn2,AlCu) were also found at the grain boundaries, but no indication of the size of these is given. The authors have argued that precipitates in the weld nugget have gone into solid solution during welding, and have been re-precipitated during cooling. This argument is reinforced by the observation that other unidentified precipitate types of approximately 10nm (intragranular) and 30-40nm (intergranular), both of which were reported in the parent material, are not observed. The former are reported to have grown to about 20nm diameter in the thermomechanically affected region of the weld. However, the time available for a complete solution heat treatment is very small, and it seems more likely that only partial dissolution will have occurred.

Typical TEM micrographs (Fig.8), show the fine and coarse precipitate in the HAZ, and the coarser precipitate in the nugget. Ellis and Strangwood [5] have also reported a slight coarsening of some unidentified precipitates in the nugget of a friction stir weld in a rapidly solidified 8009 alloy.

Further evidence for some solution treatment occurring in the nugget region can be seen by examining hardness profiles for friction stir welds in a 2014-T6 condition, as shown in Fig.3b. In the as-welded condition, the typical minimaare observed either side of the nugget. After an additional ageing treatment (18 hours at 160 degrees C) it is noted that the hardness level at the minimum in the thermomechanically affected region decreases further, implying overageing, whereas the hardness in the nugget region increases. This implies that, in the nugget region, at least some of the precipitates were taken into solution during the welding process, and probably re-precipitated in some form on cooling. Further ageing may have coarsened these precipitates to give a higher hardness level. Confirmation of this by TEM will be necessary.

Much useful information can be gained from studying joints in dissimilar metals. Figure 9 shows micrographs of a weld between 2219-T87 and 6061-T6 alloys.

Figure 9a shows the boundary between the 2219-T87 material and the nugget, and it is immediately apparent that although very high shear strains have been experienced by the materials, mixing is far from complete. However, bonding between the two alloys is clearly complete. A higher magnification micrograph in Fig.9b demonstrates these points very well. Some diffusion across such interfaces must occur, although at present no studies are known which have attempted to measure or calculate its extent.

Implications for future developments

Although very few studies have been undertaken, it is evident that the microstructural aspects of friction stir welds are very complex. To improve the weld properties, it will be necessary to control the thermal cycle, in particular to reduce the overageing effect in thermomechanically affected regions of the weld where the lowest hardness and strength are found when welding fully hardened alloys. For optimum properties, it would seem that welding prior to heat treatment is the best choice, although it is recognised that this will not be a practical solution for many applications. However, in the immediate future there is a clear need to characterise the microstructures of these welds in great detail, so that a fuller appreciation of the phenomena which occur can be achieved.

Acknowledgements

The author is indebted to the Sponsors of GSP 5651 for permission to publish part of this work, to Murray Mahoney of Rockwell International for permission to reproduce Fig.8, and Gil Sylva of Lockheed Martin Astronautics for permission to reproduce Fig.9.

References

TWI is planning further development of friction stir welding of various alloy systems. Some projects are already underway. Contact Dave Nicholas or Philip Threadgill for further details.