Technical Note: A Major Loss in Tensile Strength of Friction Stir Welded Aluminum Alloy Joints Resulting from Atmospheric Corrosion

In 1991, Thomas and coworkers¹  pioneered friction stir welding (FSW) as an emerging solid-state joining technology. Currently, FSW technology is widely used for welding many types of metallic materials, especially aluminum alloys.^1-2  The method is a hot-shear, solid-state type of welding and does not melt the material, as opposed to conventional fusion welding techniques.²  Hence, FSW shows little/no risk of cracking and/or porosity in the weld. The formation of defects in FSW specimens and the optimization of process parameters for minimizing defect formation has been comprehensively studied during the last decade.^3-6  It has been reported that the occurrence of joint line remnants (JLRs), also known as kissing bonds (KBs), in the stir zone (SZ) of the joints is almost unavoidable in mass production of Al alloy components. JLRs, which appear as zigzag lines in etched cross sections of SZs, are formed during solid-state bonding. JLRs are defined as solid particles which are in contact with their surrounding areas, but with little or no strong bonding present.² 

Okamura, et al.,⁷  were the first to study the effect of JLRs on the on mechanical properties of friction stir (FS) welded Al alloys. They reported that JLRs have no effect or only a minor effect on the tensile properties of the joints. This has been confirmed by many others, who report that JLRs do not deteriorate the tensile strength of the weldments, see e.g., Kadlec, et al.,⁸  and Le Jolu, et al.⁹  However, JLRs have been shown to have “undesired” effects on the fatigue performance of FS welds. Thus, they are reported to be preferential sites for fatigue crack initiation, hence reducing the joints’ fatigue life.^8-9  Indeed, because JLRs can hardly be detected by existing nondestructive examinations, they are a concern for FSW production/design engineers and limit more widespread adoption of FSW for critical components. It should be mentioned that other methods, e.g., nonlinear ultrasonic¹⁰  and infrared thermography,¹¹  have recently been shown to be promising for detection of JLRs in FS welds. Concerning the origin of JLRs, Reynolds¹²  initially suggested that they are remnants of the oxide layer, which was initially present on the surface of the Al sheets. Using focused ion beam milling-assisted transmission electron microscopy, Sato, et al.,¹³  confirmed Reynolds’ suggestion,¹²  showing that JLRs contain fine, amorphous Al₂O₃ particles.

Knowledge of the corrosion behavior of FS welds is necessary for many applications of FSW technology, and there have been a large number of studies on the corrosion behavior of FS welds, see e.g., Paglia and Buchheit.^14-15  Another example is the work on the corrosion behavior of FSW Al alloy joints by Frankel and Xia.¹⁶  They studied pitting corrosion and stress corrosion cracking (SCC) of FS welded and gas tungsten arc welded Al 5454-O (UNS A95454⁽¹⁾) and 5454-H34 specimens. They noticed that the “remnant boundary defects,” or JLRs, are potential concerns in terms of corrosion susceptibility in NaCl(aq). While they¹⁶  reported an appreciable decrease in weld ductility, they did not observe a considerable reduction in the tensile strength of welded specimens resulting from corrosion.

Despite the numerous studies conducted to understand the role of defects for the mechanical properties, especially the fatigue performance, of FS welds,^2,8-9,12  there is little known concerning atmospheric corrosion behavior of Al alloys joints produced by FSW. Recently, Esmaily, et al.,^17-18  investigated the NaCl-induced atmospheric corrosion mechanisms of FS welds produced using different types of tools (bobbin tool and conventional fixed pin) and different welding parameters, and noticed that pitting corrosion in the heat-affected zone (HAZ) is the most pronounced type of environmental degradation in the welds. Apart from the role of process parameters and configuration of the welding tool, no information is available regarding the possible effect(s) of FSW-related defects on the environmental degradation of Al alloy joints. Also, reports on the corrosion behavior of common FSW defects, especially JLRs, in atmospheric environments are scarce. The present paper investigates the corrosion behavior of JLRs in ambient air in the presence of sodium chloride (95% relative humidity [RH], 22°C, 70 μg/cm² NaCl, and three different concentrations of CO₂), and the consequences on the tensile strength of the joints. The tensile strength of the joints was analyzed before and after corrosion exposure, focusing on the role played by the JLRs.

The base material (BM) consisted of 10-mm-thick extruded AA6005-T6 (UNS A96005) plates. The chemical composition of the BM, analyzed by means of optical mass spectroscopy, was 0.55%Mg, 0.45%Mn, 0.19%Fe, 0.55%Si, 0.25%Cu, 0.13%Cr, 0.08%Ti, and balance Al (in wt%). The profiles were cut to the sample dimensions 300 × 250 × 10 mm³.

The weldments were analyzed using visual inspection and, in some cases, radiographic inspections, enabling characterization of the typical defects formed in FS welds. In total, 110 welding runs were performed in order to identify the optimized process parameters for welding the 10-mm-thick AA6005-T6 profiles. Joint efficiency (JE) of the weldments was also evaluated. It may be noted that this value is a numerical value, which signifies a percentage (strength of weld/strength of BM). A solution of 1.2 mL hydrofluoric acid (HF) + 1.8 mL hydrochloric acid (HCl) + 2 mL nitric acid (HNO₃) + 95 mL water was used to reveal the macrostructure of the welds.

Tensile specimens were prepared from the welded specimens (see below) and exposed to environmental degradation in an apparatus that is entirely made of glass and PTFE. The samples were suspended by a nylon string in individual chambers. The exposures were conducted in purified air at ambient temperature (22.00±0.3°C). CO₂ was added from a cylinder to give a constant concentration of 400, 800, and 1,200 (±10) ppm, i.e., three different types of exposures. RH was regulated to 95±0.3% by mixing measured amounts of dry air and 100% RH air. Gas flow was 1,000 mL/min. For details of the corrosion product removal procedure, see previous papers.^19-23 

Prior to the corrosion exposures, the samples were ground using ultrapure water on SiC papers (from P500 to P4000 mesh), followed by 3 μm and 1 μm diamond paste. Afterward, fine polishing was performed with a suspension of colloidal silica in water for 160 s. Subsequently, the specimens were contaminated with 70 μg/cm² NaCl by spraying a solution of 20 mL distilled water, 80 mL ethanol, and 1 g NaCl. At the end of the exposures and after removing the corrosion products, corrosion pits were investigated quantitatively by means of an interference microscope (RST Plus^†) using vertical scanning interferometry (VSI) mode. The interference microscope is a white-light scanning instrument with a measurement range of 0.5 mm. The instrument had a depth resolution of about 10 nm and a lateral resolution of about 40 nm.

Twenty sub-sized tensile test specimens were prepared from the weldments with optimized mechanical properties according to ASTM Standard E-8/E8M-11 sub-size specifications,²¹  using an electrical discharge cutting machine (see Figure 1). The tensile strength of the specimens was examined before and after the exposures using a constant strain rate of 8 × 10⁻⁴ s⁻¹ with a computer-controlled machine at ambient temperature. An overall inspection of the longitudinal cross sections was done using a stereomicroscope. An environmental scanning electron microscope (ESEM) was used for morphological inspections and for studying the fracture surfaces.

The first step was to find the optimized set of welding parameters, which result in high-quality weldments. Figure 2 shows the process window for the 10-mm-thick AA6005-T6. In FSW, defects can form as a result of improper flow of metal and/or insufficient heat input/consolidation of metal in the SZ.^24-26  Thus, too high and too low peak temperatures have to be avoided. In this study, FS welds that were free from defects, including porosity, tunneling phenomena, cracks, and significant flash (material being extruded or expelled from the interface), were considered as “sound welds” (see Figure 2). Also, by analyzing the JE of the welds, regions of enhanced joint efficiency could be delineated within the “sound weld” regions (see Figure 2). In this way, two sets of parameters (designated “slow” and “fast,” see Figure 2) were identified that resulted in sound weldments with enhanced JE. The slow process (with a JE of 67%) used a travel speed of 500 mm/min and rotational speed of 700 rpm, while the fast process (with a JE of 75%) was performed with a travel speed of 1,000 mm/min and a rotational speed of 1,100 rpm (the JE values were obtained from the tensile tests). Thus, while the BM had an ultimate strength (UTS) of 285 MPa, the corresponding value for the slow and fast weld were 190 MPa and 215 MPa, respectively. These two weldments were selected for corrosion and microstructural investigations.

Figure 3 shows the macrostructure of the slow (Figure 3[a]) and the fast (Figure 3[b]) FSW specimens. As seen, the welds were completely consolidated. In both cases, JLRs appeared as semi-continuous bands from the top to the bottom of the SZs. The SZ was larger in the slow FSW than in the fast FSW, meaning that more material was stirred and affected by the welding procedure (compare Figures 3[c] and [d]). Also, the area fraction of JLRs in the SZ was lower in the slow FSW (~0.3%) than in the fast one (~0.48%). A similar relation between the fraction of JLRs and welding parameters has been reported by Sato, et al.²²  In addition, image analysis showed the JLRs were not only larger in the SZ of the fast welds than the slow ones (with an average diameter of 9.5 μm and 4.8 μm, respectively), but also more abundant.

Figure 4 illustrates the topography of a typical area containing JLRs. Before the topographical study, the exact location of a JLR was found using SEM and marked using nanoindentation. Figure 4(a) shows a 3D profile of such a region, while Figures 4(b) and (c) present 2D profiles along lines parallel to the X and Y directions as indicated in Figure 4(a). A region of interest (ROI), was selected that contained the JLR and the neighboring area (indicated by a rectangle in Figure 4[a]). Figures 4(b) and (c) reveal a depression in the ROI with a maximum depth of ~5.5 μm. Analyzing several regions containing JLRs by the same methodology showed that such depressions regularly occurred in the vicinity of JLRs. It should be added here that the depressions at the JLRs were also present directly after cutting the specimens, before grinding and polishing. Figure 3(d) shows that the depth of the depressions increases sharply with the diameter of the neighboring JLR. The presence of such depressions around the JLRs in Al alloys FS welds has not been reported elsewhere.

Half of the sub-sized tensile specimens were exposed to humid air in the presence of NaCl for 100 h (95% RH, 22°C, 400 ppm CO₂, and 70 μg/cm² NaCl). Analysis of the samples before and after the exposure (see representative micrographs in Figures 5[a] and [b]) in conjunction with post-exposure analysis of the corrosion pits (Figure 6) demonstrated that the regions containing the JLRs were more heavily attacked than the other parts of the SZ, featuring much deeper corrosion pits.

Figure 6 summarizes the results on the pitting of the stir zone in FS welds. The quantitative results highlight the following main findings: (i) in both types of welds, the corrosion pits were much deeper in the vicinity of the JLRs compared to the other parts of the SZ, and (ii) the corrosion pits close to the JLRs were shallower and fewer in the slow weld (Figures 6[c] and [d]) than in the fast weld (Figures 6[a] and [b]).

Additional exposures were also performed at higher CO₂ concentrations. Thus, specimens were subjected to corrosion not only at 400 ppm, but also at 800 ppm and 1,200 ppm CO₂. After removing the corrosion products, the corroded surfaces were analyzed, focusing on pit distribution and pit depth (compare Figure 6). Figure 7 shows the depth of pits both in the vicinity of the JLRs and in the surrounding SZ as a function of CO₂ concentration. Similar to the situation at 400 ppm CO₂ (Figure 6), the pits formed at 800 ppm and 1,200 ppm CO₂ were deeper in the vicinity of JLRs than in the SZ. While an increased CO₂ concentration resulted in more shallow corrosion pits away from the JLRs, it had no influence on the depth of the pits formed in vicinity of JLRs. Note that the average corrosion rate (measured as metal loss) also decreased with increasing CO₂ concentration. The average corrosion rate measurements are not presented here as they are outside the scope of this paper.

The tensile strength of the specimens was examined after the conclusion of the corrosion exposures. Note that the tensile tests were performed before corrosion product removal. The results were compared to those of the unexposed specimens (Table 1).

The corresponding values related to the BM are also included in Table 1 (see beginning of discussion concerning the tensile strength of the BM and the fast and slow welds). Table 1 clearly shows that exposure to humid air in the presence of NaCl resulted in a significant degradation of the tensile properties of the welds. Also, the degradation of properties was more pronounced for the fast FSW specimens than for the slow welds. Thus, the UTS of the slow weld decreased from 190 MPa to 175 MPa after corrosion in the presence of 400 ppm CO₂ (a loss factor of 13%), while the corresponding value decreased from 215 MPa to 148 MPa for the fast welds (a loss factor of ~31%). It is also interesting to note that when the CO₂ level increased to 800 ppm, loss factor was even higher with a value of ~37% for the fast FS welded specimens. A similar trend was observed in the elongation values. To the best of the authors’ knowledge, there are no earlier reports of a significant decrease in the tensile strength and elongation of FS welded Al joints as a result of atmospheric corrosion. The significant deterioration of the mechanical properties of the welds called for further investigations to find out whether the observed behavior is influenced by the JLRs.

An inspection of the fracture showed that the unexposed specimens fractured in the HAZ, which is typical for a healthy weld.^2,27  In contrast, all of the exposed specimens fractured in the SZ, following the zigzag path of the JLRs. Figure 8 shows representative SEM micrographs indicating that the fracture surface of the unexposed samples is dominated by fine dimples with few voids, signifying a ductile fracture mechanism. In contrast, the samples subjected to atmospheric corrosion exhibited fracture surfaces that were associated with JLRs in the SZ (Figures 5 through 7). Also, the fracture surface of the exposed specimens was composed of the dimple structure with an abrupt transition to what appears to be a free surface in the JLRs. It may be noted that Frankel and Xia¹⁶  (FSW sulfide stress cracking) reported a similar structure/morphology of the fracture surfaces in Al alloy FS welds.

Under the experimental conditions (air at 22°C and 95% RH), the NaCl deposited before exposure absorbs water to form an aqueous solution on the sample surface. Hence, the atmospheric corrosion investigated can be regarded as a special case of corrosion in aqueous solution. According to numerous researchers, see e.g., Birbilis, et al.,^28-29  and Ilevbare, et al.,³⁰  Al suffers electrochemical corrosion in aqueous solution. In neutral and alkaline solution, the anodic dissolution of Al is reported to occur by the two coupled anodic reactions (Reactions [1] and [2]), while the cathodic reactions are hydrogen evolution (Reaction [3]) or oxygen reduction (Reaction [4]):

The corrosion behavior of Al and its alloys is dominated by the passive oxide/hydroxide film, which protects the metal surface in most environments.^29-30  Hence, Al usually suffers pitting corrosion when exposed to the ambient atmosphere, reflecting local failure of the passive film.²⁹  Al alloys have also been reported to suffer crevice corrosion in the atmosphere.³¹ 

The passive film formed on Al alloys is not fully homogenous but contains flaws associated with, e.g., second phase particles and grain boundaries. Noble second phase particles are favorable cathodic sites, the corresponding anodic dissolution of Al causing pitting of the surrounding matrix. Thus, the corrosion properties of Al alloys are strongly affected by the presence of various types of secondary phases in the alloy’s microstructure, including precipitates and dispersoids.^28-31 

Considering the crucial importance of microstructure (especially second phase particle characteristics) for the corrosion of Al alloys, it is tempting to explain the corrosion behavior of Al alloy FS weldments in terms of FSW process-induced microstructural changes in the alloy. It is well established that the SZ in Al alloys experiences temperatures in the range 450°C to 570°C.^32-35  Also, the microstructure of Al alloys is known to be sensitive to the thermomechanical deformation occurring during FSW.³²  Consequently, the FSW process has profound implications for the microstructure of, e.g., Al-Mg-Si alloys. Thus, it is reported that while strengthening particles (Mg₂Si) are dissolved in the SZ of the FSW specimens, large (1 μm to 12 μm) Fe- and Si-rich particles are not dissolved but become fragmented, forming much greater numbers of smaller particles compared to the BM.^36-37  It has been shown by several authors that Fe- and Si-rich particles (e.g., Al₈Fe₂Si, Al₅FeSi, Al₁₂Fe₃Si, and Al₁₂(FeMn)₃Si) are abundant in the microstructure of the SZ in 6000 series Al alloys FS welds.^36-38  Thus, the pit formation in the SZ (excluding the JLRs) of the specimens (Figures 5 and 6) is probably linked to the presence of such second phase particles.

The mechanism behind the corrosiveness of chloride toward Al has been discussed by Frankel,³⁹  Kruger,⁴⁰  and more recently by Strehblow.⁴¹  It is well established that chloride ions destabilize the alumina passive film and cause pitting corrosion of Al alloys. In addition, soluble chlorides form aqueous electrolytes that can support electrochemical corrosion. Consequently, the atmospheric corrosion of Al is strongly accelerated by NaCl.^31,39-41 

The NaCl-induced atmospheric corrosion of several metallic materials, including zinc,^42-43  magnesium,^20,24  and Al,^44-45  is reported to be inhibited by CO₂. This is mainly an effect of its acidity. Thus, CO₂ dissolves in the aqueous surface layer forming carbonic acid with pk_a1 = 6.35 and pk_a2 = 10.33 (Equations [5] and [6]):^20,24,44-45 

The inhibitive effect of CO₂ on the corrosion of Al in humid air in the presence of NaCl has been attributed to neutralization of the cathodic alkalinity.⁴⁴  Thus, the NaOH(aq) solution generated at the cathodic sites by the corrosion process (Reaction [4] plus migration of Na⁺(aq) to the cathodes) is neutralized by CO₂, i.e., Reactions (7) and (8). The corrosion process will be affected by the resulting pH decrease in several ways. First, alumina becomes less soluble, causing precipitation of Al hydroxide close to or even on the cathodic sites. It is proposed that the resulting alumina precipitate will tend to slow down the cathodic process by restricting the access of reactants to the electrode surface. In addition, replacement of hydroxide ions by carbonate ions in the surface electrolyte is expected to decrease surface conductivity. Both effects will tend to impede electrochemical corrosion. This effect is suggested to explain the tendency for the pits in the SZ of the welds (away from the JLRs) to become shallower with increasing CO₂ concentration (Figure 7).

All exposed specimens fractured in the region that contain the JLRs, where a ductile/cleavage mechanism was observed (Figure 8[b]). This implies that fracture was preceded by extensive degradation of the mechanical properties of the weldments in the vicinity of the JLRs. Indeed, the results strongly suggest that the degradation of mechanical properties is connected to the severe pitting observed at the JLRs (see Figures 5 through 7). Thus, it becomes important to understand: (a) why the JLRs are preferentially attacked by pitting, and (b) how corrosion affects the tensile strength of the welds.

As mentioned earlier, pitting corrosion of Al alloys tends to involve microgalvanic corrosion cells involving cathodically active second phase particles. The Al₂O₃ particles that are aggregated in the JLRs (see above) are electronic insulators. Hence, they are not cathodically active and can play no active role in pitting. Instead, the depressions observed in the specimens in the vicinity of the JLRs are suggested to be significant in this context (see Figures 4[a] through [d]). The observation of these depressions directly after cutting proves that they are not a result of, e.g., pull-out of oxide particles during grinding and polishing, implying that they are either an inherent feature of the FSW weld or a result of the cutting of the welded specimens. Indeed, considering that cutting does give rise to significant tensile forces (Figure 9[a]), the depressions in the specimen surface (depth depending on the size of the JLRs, Figure 4[d]) may have formed during cutting.

The observation that the depth of the corrosion damage was independent of CO₂ concentration (Figure 7) suggests that the preferential corrosion of the areas containing the JLRs is linked to crevice-type corrosion. In this scenario, the depressions appearing at the JLRs correspond to micrometer-sized (and presumably sub-micrometer sized) cavities within the JLRs. The surface electrolyte penetrates into these cavities, causing corrosion to occur in an environment that is more or less separated from the sample surface. The occluded volume would be little influenced by the neutralizing effect of carbonic acid, causing the electrolyte-filled cavities be unaffected by the corrosion-inhibitive effect of CO₂. This explains both the deep pitting in this region and the lack of a CO₂ concentration effect on corrosion in the JLRs. Oxygen is also expected to be depleted in the occluded volume. If O₂ depletion is deep enough, the cathodic reduction of oxygen will be replaced by water reduction (Reaction [3]).The cathodic alkalinity produced in the occluded volume would tend to keep the cathodic sites active, explaining why the JLRs become preferentially corroded.

Corrosion affected the mechanical properties of the welds to a large extent (Table 1). This undoubtedly means that corrosion pits in the region containing the JLRs acted as precursors to cracking and, accordingly, to failure during the subsequent tensile testing. The authors believe that there are three possible reasons for the observed deterioration of the welds’ tensile properties resulting from corrosion (see Figure 9[b]).

First, as mentioned above, the electrolyte-filled cavities are considered to be unaffected by the corrosion-inhibitive effect of CO₂, explaining the deep pitting in this region. The results thus imply that the significant corrosion-induced decrease in the tensile strength and ductility of FS welds affecting the region containing the JLRs is caused by crevice-type corrosion associated with the narrow and relatively deep surface cavities observed at the JLRs (see Figure 4).

Second, SZ of FS welds contains tensile residual stresses (RSs). Thus, several researchers have reported^46-48  that the tensile RS in the weld area of this family of Al alloys can reach 50% to 60% of the yield stress. RSs in FSW specimens are caused by interactions between the thermal history (time and temperature), deformation (stress and strain), and the material microstructure.⁴⁹  It was suggested earlier that the cavities observed at the JLRs become filled with electrolyte, causing crevice-like corrosion. Thus, the combination of RSs and the corrosive environment in the cavities is suggested to have resulted in SCC of the weld specimens during the atmospheric corrosion exposure.

Third, water reduction occurs within O₂ depleted cavities, causing hydrogen pickup by the metal⁵⁰  (above), with the dissolved hydrogen further deteriorating the tensile properties of welds. Hydrogen pickup may be accelerated by significant RSs that are present in the SZ of FS welds. The sharp transition from the ductile structure to the free surfaces (or brittle faces) observed in the fracture surfaces in the vicinity of the JLRs (Figure 8[b]) is an indication of hydrogen pickup. Hydrogen is reported⁵¹  to be attracted to regions of high triaxial tensile stress, in this case in the grain boundaries adjacent to the deep pits in the JLR region. Thus, hydrogen picked up by the alloy during the corrosion has assisted in the fracture of the specimens, possibly by making cleavage easier and/or by assisting in the development of intense local plastic deformation.

The observations therefore imply that the combination of enhanced pitting (CO₂ depletion in the cavities around the JLRs), SCC (during atmospheric corrosion exposure), and hydrogen damage (acting during straining the welds) give rise to the significant decrease in the tensile strength and ductility of the FS welds. However, the results at hand do not allow one to conclude on the relative importance of the different factors.

It may be noted that while the unexposed fast welds exhibit a higher JE (or higher UTS) than the slow welds, the trend in strength of the weldments is reversed after a short exposure to NaCl-induced atmospheric corrosion. Based on the earlier discussion, the more pronounced deterioration of the mechanical properties of the fast FS welds compared to the slow welds (see Table 1) is related to the tendency of the JLRs to be larger in the former case, considering the observations showing that cavities become deeper when the JLRs are larger (Figure 4[d]). Also, the more intense corrosion observed in the vicinity of the JLRs in the fast weld is attributed to the size and distribution of the JLRs (see Figures 6[b] and [d]). The results strongly suggest that attention should be given to minimize the area fraction and size of the JLRs through optimization the FSW process parameters, as cutting of the welded specimens, which was suggested to be the reason for the formation of JLRs-induced cavities, is an unavoidable part of manufacturing of welded structures.

Also, the results provide new insights into the influence of a common FSW defect on the universal properties of the welds. The discovery of the role played by JLRs in environmental failure of FS welds is of technological importance for FSW users as the welds are normally subjected to both atmospheric corrosion and mechanical load. While the JLRs have previously been considered to have no harmful effect on the tensile strength of FS welds, the present paper shows that they have a strong impact on the tensile strength and elongation of welds after a (very) short period of atmospheric corrosion exposure.

This paper investigates the effect of atmospheric corrosion on the tensile properties of a FS welded Al 6005 alloy. It was shown that:

• The areas containing the JLRs are clearly preferentially corroded compared to the other parts of the SZs. The rapid corrosion is attributed to crevice-type corrosion occurring as a result of a limited access of CO₂ as consequence of cavities present around the JLRs. The presence of cavities/depressions around JLRs in the SZ of the welds are evidenced in 3D and 2D topographical images.

• Tensile tests performed before and after the corrosion exposures showed that strength of the welds decreased significantly (in some cases by ~37%) as a result of corrosion. Inspecting the fracture surface of the welds indicated that the regions containing the JLRs had formed deep corrosion pits and that the loss of tensile properties was a result of corrosion. This is explained by a mechanism involving hydrogen pickup, a special type of SCC associated with static (i.e., residual) stresses, and also the deep corrosion pits formed as a result of the limited access of CO₂ by the metallic substrate during atmospheric corrosion.

• Deterioration of the tensile strength was more serious in fast welds than in slow ones. This is suggested to be linked to a higher area fraction and larger size of JLRs, which leads to the formation of deeper cavities in the vicinity of the JLRs in the former case.

The authors would like to deeply thank the Swedish Foundation for Strategic Research (SSF) (grant number: RMA08-0138) and the Swedish Research Council (Vr) (grant number: 2015-04977) for funding this research. ME sincerely thanks Helge AX:son Johnsons Stiftelse and ÅF research grant for financially supporting some of the experiments.