Cracking in the heat affected zone (HAZ) of Al-Mg-Si alloys - a review

TWI Bulletin, September/October 1990

Mike Gittos is a Principal Metallurgist in the Materials Department. He graduated from Manchester University, joining the then Metallurgy Department at TWI in 1974. Although most of his work has been concerned with the welding of non-ferrous metals, his published work also relates to ferritic and stainless steels as well as plastics materials.

Mike Scott graduated from Cambridge with a degree in natural sciences, specialising in metallurgy. In his first post, with the Royal Naval Scientific Service, he worked on electroplating applied to manufacturing radar valves and ancillary devices, and on the manufacture of ceramic ferrite.

From there he moved to the Central Electricity Research Laboratory where he worked on the metallurgy of magnesium alloys used for sheathing uranium fuel elements for nuclear power stations. In 1960, he joined the then British Welding Research Association, where he spent many years as leader of the Non-Ferrous Group in the Materials Department. He is now a consultant within that Department. During his time at Abington, he has worked on various aspects of joining most of the common non-ferrous metals and their alloys and has been much involved in microjoining.

The problem of HAZ cracking in Al-Mg-Si alloys was first described in a study which was part of TWI's Co-operative Research Programme. This article by Mike Gittos and Mike Scott reviews fundamental crack mechanisms.

Al-Mg-Si alloys are medium-strength Al alloys which offer excellent corrosion resistance and extrudability. They are also the only heat-treatable Al alloys which are widely fusion-welded. If the alloys are to be used satisfactorily in welded structures, the locally-reduced strength zone at welds must be taken into account, and welding procedures must be adopted which prevent weld metal and HAZ cracking.

Since publication of a TWI investigation of the latter problem, ^[1,2] two groups of workers have published the results of HAZ cracking studies on similar alloys ^[3,6] and further cases have continued to come to TWI's attention. In addition, a re-examination of earlier published work ^[7] has revealed further evidence of the phenomenon, and HAZ cracks have been observed during a study of weld metal cracking in 6061 Al. ^[8]

Several different crack mechanisms have been proposed, so it seems appropriate to review the problem of HAZ cracking in Al-Mg-Si alloys in the light of recent additional information.

The story so far

TWI work ^[1,2]

Circular patches of a range of diameters were welded into 3mm thickness 6082 sheet using the TIG process. This arrangement simulated the weld geometry where cracking had been observed. It was also intended to pursue an observation that cracking was prevented when the effective patch diameter was increased.

However, no relationship with patch diameter was established, though all patches welded with Al-Mg filler metal cracked whereas all patches welded with Al-Si filler metal did not. It was suggested that weld metal solidus was the governing parameter, a hypothesis supported both by the observed effect of a variation in filler metal dilution (and hence also the weld metal solidus) in tests using Al-Mg filler and by the field experience of cracking: the problem was normally associated with the use of Al-Mg fillers in situations where the dilution by parent metal was likely to be high.

Japanese work ^[3,4]

Tanaka and his co-workers used MIG welding to make straight butt joints in 3mm thickness sheets of 12 Al-Mg-Si alloys. No major cracking was encountered in this work and no cracking parallel to the weld (the case of most practical concern) apparently developed. The welds were assessed in terms of the degree of melting (liquation) in the HAZs; this was more extensive in welds made with Al-Si filler metal and in parent metals of higher Si content.

Differential thermal analyses (DTA) of some welds were made and it was shown that minor cracks were present only in the weld in which the solidus temperatures of weld metal and parent metal were closest (6082 welded with Al-Mg filler). Some loss in tensile strength was reported for the welds in two of the higher Si-bearing parent metals made with Al-Mg filler metal: this could have been due to the presence of cracks but there was no report of examination of the broken tensile testpieces.

From the reported microanalyses of melted grain boundaries in HAZs, it seemed that partial 'back-filling' with liquid from the molten pools had occurred.

Tanaka et al disagreed with the cracking mechanism proposed by the authors on the basis of their DTA results. (Data presented for three of the parent metals indicated that their solidus temperatures were higher than those of the weld metals made with both Al-Si and Al-Mg fillers.) They proposed that Mg and Si invade the HAZ, either by solid-state diffusion or as liquid, forming low-melting-point constituents.

Canadian work ^[5,6]

Katoh and Kerr used MIG and TIG welding for Varestraint tests in 6061 (including modifications with 0-1%Cu), 6261 and 6531 alloys of 6.3mm thickness.

In MIG welds, two types of cracking were identified: longitudinal, similar in morphology to those associated with conventional welds, and transverse. The TIG welds, which were autogenous, caused transverse HAZ cracks only. It was suggested that the transverse cracks were formed by a type of liquid metal penetration mechanism which was exacerbated by Al-Si filler metal (in the case of MIG welds). The longitudinal cracks were more consistent with cracking which has been more widely observed: in addition to their orientation with respect to the weld, they were produced with Al-Mg but not Al-Si filler and with Al-Mg they formed with the application of very little strain.

Thermal analyses demonstrated that, for 6061, the solidus of weld metal made with 5356 (Al-Mg) was closer to that of the parent metal, but that the solidus of Al-Si weld metal was much lower. However, just as Tsujimoto et al ^[3] had shown, the parent metal solidus was slightly higher than the Al-Mg weld metal solidus (by 13 and 10°C in the two investigations).

Katoh and Kerr also discounted the authors' model on the basis of their DTA results but went on to propose one which was essentially similar. The precursor for the operation of this model was to invoke socalled 'constitutional' liquation ^[9] to account for a lowering of the temperature at the onset of melting in the HAZ compared with that measured by DTA. Thus the DTA did not describe events in the weld accurately because either the heating and/or cooling rates were inappropriate to welding or the technique was not sufficiently sensitive to detect the salient local melting or solidification at grain boundaries. (A comparison of the DTA and hot tensile test results obtained by Tanaka et al ^[4] indicates a similar disparity.)

Other cases of HAZ cracking

Since TWI's work was reported, there have been several further cases of HAZ cracking. All documented cases from fabricators have involved the use of Al-Mg filler except for one notable incident in which pure Al filler metal was inadvertently used for welding 6082 Al.

HAZ cracking has also been observed in a study of cracking in Al-Si weld metal in 12mm thickness 6061 using the TIG process for Varestraint testing ^[8] For autogenous TIG welds, HAZ cracks were observed perpendicular to the rolling direction of the sheet. Thus cracking was either parallel or perpendicular to welds, depending on the orientation of the test specimen. In common with the longitudinal cracks which Katoh and Kerr found in MIG welds, ^[5] the HAZ cracks were connected to weld metal cracks.

An examination of the work reported by Houldcroft on weld metal cracking in various Al alloys ^[7] shows that examples of the Houldcroft test using H30 (6082) Al produced cracks both along the centreline and at the edge of the weld when NG6 (Al-Mg) filler was used. With NG21 (Al-Si) filler, the cracking was restricted to the centreline. Although no sectioning of these tests was reported, it may be surmised that this edge-cracking was in fact HAZ cracking.

Discussion

Although both investigations subsequent to the authors' work concluded that DTA results disproved the original cracking model, a critical examination of the two sets of results does not support this view. The DTA results obtained in both studies showed that, according to the authors' model, HAZ cracking would not occur: this was borne out by the experimental results. Tsujimoto et al ^[3] observed greater HAZ liquation with parent metals in which Si was relatively high and with Al-Si filler. However, it would appear that small cracks were observed in only two of the welds and none of the sections which was assessed reached the category of 'minor HAZ cracking'.

Similarly, the MIG bead-on-plate deposits made by Katoh and Kerr ^[5] using Al-Mg filler did not cause any HAZ cracking except in association with bending at the end of the test. Since the maximum equilibrium solidus of Al-Mg weld metal is 595°C, the alloys for which solidi were measured in the two investigations (6061 and 6082 with solidi of 598 and 597°C) would be resistant to HAZ cracking at all levels of dilution according to the authors' model.

Evidence from Houldcroft's work is also supportive of the authors' model. The change in crack location during the progress of the test could reflect the high sensitivity to dilution demonstrated subsequently. ^[1] The failure to identify HAZ cracking separately during the following 20 years or so was probably due to the fact that the remedies for both weld metal solidification cracking and HAZ liquation cracking in these alloys are similar, ie both are tackled by avoiding high dilutions and Al-Si proves to be the more crack-resistant filler for both mechanisms.

The case of HAZ cracking associated with the use of pure Al filler metal is also consistent with the hypothesis that high weld metal solidi are likely to cause cracking. Indeed, use of a pure Al filler wire would be a good test of the authors' model: it predicts that such a consumable is likely to cause HAZ cracking when used for restrained welds in all 6XXX alloys, irrespective of dilution. Examination of the weld metal solidi (see Figure 1), shows that the equilibrium weld metal solidi will always exceed those of the parent alloys.

The Varestraint test imposes extreme conditions of strain not fully reproducing the practical situation, and the cracking produced in 6XXX Al alloys does not seem representative of cracking in conventional welds. The crack morphologies which have been reported are summarised in Figure 2. In the work described by Katoh and Kerr ^[5,6] several features of the cracks - orientation (transverse cracks), increased severity with Al-Si filler metal, and continuity with weld metal cracks - were not typical of those observed in real welds.

Although some of the longitudinal cracks in Varestraint coupons made with Al-Mg (MIG) and Al-Si (TIG) were superficially similar to cracks in conventional welds, these also had peculiar features. In particular, they extended a relatively long distance behind the molten pool, occurred when Al-Si filler metal was used, and were invariably continuous with weld metal cracks. Since they did not occur until the coupons were bent, cracking must have taken place well below equilibrium solidi for both weld and parent metals. From the data for the peak temperatures in the HAZ of the TIG test, ^[6] it seems probable that longitudinal cracking of the MIG weld was occurring in a region where the temperature had fallen below 500°C. (Steeper thermal gradients would normally be expected with MIG compared to TIG welding and therefore adjusting for process could imply that cracking occurred at still lower temperatures.)

It is suggested that both the strain regime applied and the existence of a path for the access of liquid metal from the weld pool to the HAZ complicate the interpretation of cracking in these samples. It is clearly difficult to relate the results from these Varestraint tests to practical problems.

An obvious criticism of the authors' model is the simplifying assumption of equilibrium solidification. In practice, a qualitative consideration of the trends expected in non-equilibrium solidification does not alter the basic arguments. Thus, for 6082, the lowest temperature solidification products likely in the parent metal or Si-enriched weld pools will be similar and therefore no HAZ cracking would be predicted. With Mg-enriched weld pools, the weld metal composition lies at the same side of the Mg ₂Si phase field solidus peak as the parent metal only at very high dilutions.

There will, therefore, be a tendency to higher terminal solidification temperatures in the weld pool as soon as the Mg concentration is high enough to pass the Mg ₂Si quasi-binary line. As the Mg content of the pool increases further, there is the possibility of even lower melting material than in Si-enriched compositions; this explains the satisfactory performance of Al-Mg fillers at lower dilutions.

The main effect of incorporating non-equilibrium solidification trends into this model is to decrease the significance of variations in parent metal composition. Virtually all compositions of alloy 6082, for example, become relatively susceptible to cracking.

Summary

Investigations into HAZ cracking in 6XXX alloys have shown that a number of complex cracking phenomena can occur. Cracking at relatively low temperatures implies that solid-state cracking mechanisms may be operative as well as liquation effects at higher temperatures.

Back-filling of cracks has been observed, and this may complicate subsequent interpretation. Although more work is needed to understand the observed effects fully, liquation cracking in the partially-melted HAZ appears to be the problem which affects conventional welds, and the benefit of using Al-Si filler metals can be correlated with their lower solidification temperatures.

The adoption of Al-Si fillers is the usual practical solution to this problem.

References