Microstructure and mechanical properties of friction stir welded AA6092/ SiC metal matrix composite

There is a need for improved understanding on the e ﬀ ects of friction stir welding (FSW) on the metallurgical and mechanical properties of aluminium matrix composite (AMC). In this study, AA6092/SiC/17.5p-T6 AMC joints were produced by using FSW with varying tool rotation and traverse speeds. The microstructural character- isation by scanning electron microscopy equipped with electron backscattered di ﬀ raction (EBSD) system revealed a substantial grain re ﬁ nement and a homogeneous distribution of reinforcement particles in the nugget zone. The grain size of the nugget zone was greatly in ﬂ uenced by weld pitch, as a key indicator to control the amount of heat input, exposure time and cooling rate. Vickers microhardness pro ﬁ le across the welding zone revealed a signi ﬁ cant di ﬀ erence in microhardness among the base metal, heat a ﬀ ected zone, thermo-mechani-cally a ﬀ ected zone and nugget zone. The tensile strength of the cross-weld specimens showed a high joint e ﬃ ciency of about 75% of the base metal combined with relatively high ductility. Low-cycle fatigue properties were investigated in the axial total strain-amplitude control mode (from 0.3% to 0.5%) with = . The results indicate that the fatigue life of the cross-welded joints varies with grain size in the nugget zone and it is lower than that of the base metal. A signi ﬁ cant improvement of fatigue life is found to be related to the ﬁ ner equiaxed grains dominated by high angle grain boundaries in the nugget zone.


Introduction
Advanced materials like aluminium matrix composite (AMC) are considered to have considerable potential for lightweight applications due to their superior mechanical and physical properties at elevated and ambient temperature, compared to the unreinforced aluminium alloys [1]. AMCs reinforced with ceramic particles are more attractive than those of the other geometries of reinforcement such as flake and fibre, owing to their isotropic properties, oxidation resistance, higher operating temperature, and ease of fabrication [2]. However, the toughness of AMC in term of ductility and fracture toughness is lower than that of its matrix alloy [3]. Thus the improvement of the fatigue life of AMC is most pronounced in the high cycle fatigue (HCF) regime than in low cycle fatigue (LCF) as a result of ductility exhaustion [3][4][5]. Regarding the joining of AMC by the conventional welding process, it has been observed that there is a deleterious reaction between the molten metal and reinforcement particles with inhomogeneous redistribution of reinforcement particles frequently occurred in the fusion zone [6]. As a result, the weldability of AMC by fusion processes is limited, and the attractive mechanical properties of these materials have yet to be entirely utilised [7,8].
As a solid-state thermo-mechanical joining technique, friction stir welding (FSW) is considered a promising technique for joining AMC. In FSW, the combination of heat, and high plastic strain involved during the welding process can produce finer grain and more homogeneous distribution of the reinforcement particles in the nugget zone (NZ) [9,10]. However, FSW of AMC is not an easy task, and significant challenges remain in achieving optimum welding window and sufficient mechanical properties [11].
In recent years, work has been reported in looking into the effect of FSW parameters on the metallurgical and mechanical properties of aluminium alloy (6xxx) as a base matrix reinforced with various ceramic particulates, including, AA6061/SiC [12,13], AA6061/B 4 C [11,14], AA6061/ZrB 2 [15,16], AA6061/AlN [17], AA6061/Al 2 O 3 [18,19], and AA6063/ B 4 C [20]. Results showed that the welding parameters play a significant role to produce sound joints. Periyasamy et al. [12] reported that the tool rotation and traverse speeds influence the generated temperature in the NZ and subsequent metal flow and grain growth. Intermittent or turbulent metal flow characterises the stirring zone when insufficient or excessive heat input was used. Both of these conditions produce defective welds, as the stirring rate of plasticised material determines defects formation in the NZ. Tunnel defect is more likely to happen at an excessive stirring rate, while lack of bonding is formed at an insufficient stirring rate [21], suggesting a narrower welding window of AMC than that for monolithic aluminium alloys. Also, the strength of FSW joints is significantly affected by welding parameters. The joints efficiency (JE) (the ratio of the ultimate tensile strength of joints to the ultimate tensile strength of the base metal (BM) of AMC joints was lower than the BM [13,14,17,21], which is resulted from the formation of coarse grains, lower hardness, and presence of defects.
Although effort has been focused on the effect of FSW parameters on the microstructure and corresponding tensile and hardness properties of 6xxxAl composite joints, the fatigue behaviour of the welded joints have been examined to a limited scope. In this context, Minak et al. [22] and Pirondi et al. [23] have reported the HCF properties and fatigue crack propagation of AA6061/Al 2 O 3 /22p FSW joints. They reveal that the fatigue behaviour of welded joints was dependent on the evolved microstructure in the NZ. Ceschini et al. [24] investigated the effect of the FSW process on the LCF of AA6061/Al 2 O 3 /20p FSW joints. A reduction in the fatigue life of FSW joints as compared to that of the BM has been observed as a result of surface roughness, and the induced plastic strain during welding. More recently, Ma et al. [25] identified in their critical reviews that insufficient investigations had been carried out focusing on the LCF behaviour FSW joints, although many of the envisaged applications for the AMCs undergo cyclic loading with high amplitude of strain and stress during service life resulting in fatigue damage and severely limit the useful life and the performance of the components [26]. Furthermore, the LCF in the composite metals may occur because of repeated localised yielding near stress raisers (strengthening particles) that lead to stress concentration, despite the bulk of the component under elastic deformation. Thus it remains an open question on how to improve the strength and fatigue behaviour of AMCs joints.
To this end, no work has been done on the FSW of AA6092/SiC/ 17.5p-T6. Thus this work aims to fill this gap and make a detailed assessment of the feasibility of FSW application for this alloy. The focus of the study is to investigate the effect of two most important FSW parameters (tool rotation and traverse speeds) on the thermal history and microstructural changes, and to evaluate the resultant mechanical properties using microhardness, tensile and LCF testing. The obtained results are elucidated in the light of the interaction between the welding parameters and the existing theoretical background.

Material and experimental procedures
The as-received AMC sheet is AA6092/SiC/17.5p-T6, manufactured by powder metallurgy technique and subsequently rolled into 3.1 mm thick sheets. An aluminium alloy of AA 6092 was used as a matrix reinforced by 17.5 wt%. of SiC particles with an average size of around 6 µm. Sheets having dimensions of 110 mm × 50 mm were welded parallel to the rolling direction at TWI Technology Centre (Yorkshire) by using a bespoke FSW machine, and a total of nine joints denoted as W1-W9 was performed as shown in Table 1.
The welding tool was manufactured from AISI H13 with 6°flat edge featureless concave shoulder of 16 mm in diameter and an M6 threaded cylindrical pin having a length of 2.8 mm with a domed end. A plasma nitriding heat treatment was carried out to the tool to achieve a hardness of 950 Hv. The tool rotation axis was tilted by a 2°from the sheet normal and rotated anticlockwise, and the plunge depth was controlled to be ∼0.05 mm, which means the depth of the shoulder into the top surface of the workpiece. The maximum welding temperature in the NZ was measured by FLIR infrared camera. Eight K-type thermocouples (TC) of 1.6 mm diameter were embedded on the retreating side (RS) (TC1-TC4) and the advancing side (AS) (TC5-TC8), 10 mm away from the weld centre and 3 mm between each other to measure the thermal distribution across the joints as shown in Fig. 1(a).
For microstructure examinations, samples were cut perpendicular to the welding direction (cross-welds) ( Fig. 1(a)) then prepared according to the standard metallographic techniques for specimen preparation consisting of grinding with 220, 400, 800, 1200, 2000 and 4000 grit emery papers followed by cloth polishing with 6 and 1 µm diamond. Finally, BUEHLER vibratory polisher was used with 0.025 µm colloidal silica to obtain a high surface finish. The microstructure observations were carried out by high-resolution field emission gun scanning electron microscope (FEG-SEM; JEOL, 7100F) equipped with an Oxford electron-backscatter diffraction (EBSD) system operated at 15 kV with a step size of 0.25 µm and 70°stage tilt. A lower limit boundary misorientation angle cut off 3°was used to avoid spurious boundaries, and a 15°misorientation criterion was used to differentiate between lowangle grain boundaries (LAGBs) and high-angle grain boundaries (HAGBs).
For mechanical testing, cross-weld flat dog-bone tensile and LCF specimens were extracted from the FSW joints as shown in Fig. 1(a). In addition to this, one longitudinal tensile sample was extracted from the W3 FSW joint (WL3) ( Fig. 1(b)) to compare the influence of the tensile strength along the longitudinal and transverse direction of the weld. The samples were then ground and polished parallel to the principal load direction to exclude the effect of stress concentration, and nonhomogeneous distribution of thicknesses caused by the FSW. Tensile tests were performed on samples with a gauge width of 6 mm and length of 37 mm according to ASTM E8M-13a by an INSTRON 5581-50KN instrument at a constant strain rate of 1 × 10 −4 s −1 , with an extensometer having a gage length of 25 mm. The LCF tests were carried out by INSTRON 8801-50KN servo-hydraulic test machine on specimens with a gauge width of 8 mm and length of 28 mm according to the BS ISO 12106:2017 in a fully reversed total strain-controlled condition (R= −1). The strain function exhibited triangular waveforms and the test conducted at a constant strain rate of 5 × 10 −4 s −1 at ambient temperature with anti-buckling restraints. A strain amplitudes of 0.3-0.7% were used for the BM and 0.3-0.5% for cross-weld samples. An axial clip extensometer with a gage length of 12.5 mm was used to control the amount of applied strain, and the fatigue failure criteria were defined when the maximum tensile stress dropped by 30% below that at initial life. Vickers microhardness measurements were taken on the cross-weld by using a Buehler hardness tester with 100 gf load. An average of three hardness readings was taken at cross-weld; in the mid-thickness and above and under by 0.5 mm. Fig. 2(a) shows the effect of welding parameters on the generated peak temperature in the NZ. As the tool rotation speed increased from 1500 to 2100 rpm at a constant tool traverse speed of 25 mm/min, the generated peak temperature increased from around 500-605°C. On the other hand, the rises of traverse speed from 25 to 100 mm/min lead to a reduction of the peak temperature from about 500-488°C at a constant tool rotation speed of 1500 rpm. There is no significant change in the welding temperature, and it tends to be saturated when the rotation speed rises from 1800 to 2100 rpm, this phenomenon was also confirmed experimentally by Sato et al. [27]. This observation can be related to the reduction of the friction coefficient and torque when the weld metal softens at a high temperature, which leads to a reduction in heat generation by mechanical work and stabilises the temperature to avoid the melting of metal [28]. Fig. 2(b) compares the thermal history in the centre of NZ for W1, W3, W7, and W9 welds (corners of welding matrix). It is evident that the temperature gradient from the peak temperature of the NZ decreases with the reduction of traverse speed. The increase of traverse speed from 25 to 100 mm/min at a constant rotation speed of 1500 rpm reduces the cooling time from peak temperature from about 150 s to around 40 s, which indicates higher cooling rate [29]. Fig. 2(c) shows the typical thermal history measured in the AS and RS as a function of distance from the welding line for the joint welded at highest heat input (W7). The recorded temperature shows a nearly symmetric trend on both sides concerning the weld line. However, it is clear that the temperature at the RS (TC1-TC4) is higher than that at the AS (TC5-TC8); this behaviour is consistent with the study by Schneider [30], and can be attributed to the colder metal on the AS ahead of the tool with hot metals swept to the RS.

Metallurgical characterisation and microstructure evolution in the nugget zone
The surface morphologies of typical FSW joints fabricated by using tool rotation speeds of 1500 and 2100 rpm at constant tool traverse speed of 25 mm/min (W1 and W7) are shown in Fig. 3((a) and (b)). Visual observations revealed that a rough surface was formed at the lowest tool rotation speed of 1500 rpm ( Fig. 3(a)), which is characteristic of too cold processing conditions [31]. Smoother surface with flash, however, occurs at the highest rotational speed of 2100 rpm ( Fig. 3(b)), due to more softening of material with increasing rotational speed, in agreement with another author [31]. The typical cross-weld macrostructure of AA6092/SiC FSW joint (W5) is revealed in Fig. 3(c). It can be seen that the friction stir zone (FSZ) consists of three different zones, namely heat affected zone (HAZ), thermomechanically affected zone (TMAZ), and nugget zone (NZ) with an elliptical shape, which is identified explicitly in all joints, the same as that reported by Feng et al. [32], and Cavaliere et al. [33]. The joint exhibits a continuous flow of plasticised material between AS and RS. Under all welding conditions joints free of volume defects such as tunnels, and pinholes were formed with complete penetration. This indicates that the generated frictional heat, stirring action, and material plastic deformation are sufficient for   complete plasticization and material flow for proper coalescence [11]. However, joint W8 contained a macro defect under the top surface ( Fig. 3(d)), which is considered to be related to the fretting damage of the nitride layer.
To evaluate the effects of both of the tool rotation and traverse speeds on the distribution of SiC particles in the matrix alloy, as well as matrix grain structure, the SEM and EBSD results of the BM and typical FSW joints (W1, W3, and W7) are presented in Figs. 4 and 5, respectively. As shown in Fig. 4, the microstructure of the BM shows a reasonable distribution of the reinforcement particles in the matrix alloy having an average size of around 6 µm and an average volume fraction of 17.5%. On the other hand, the SEM micrographs of W1, W3 and W7 FSW joints show a fairly homogeneous distribution of SiC in the NZ as a result of severe plastic strain caused by the rotating tool [11]. More homogeneity was achieved at lower welding traverse speed (W1). It was also observed that the SiC particles were fractured in the NZ (white arrow) and the edges of the particles were blunted compared to the sharp SiC particles in the BM. This leads to refinement of SiC with an average size of around 1, 3, and 1.7 µm in W1, W3, and W7, respectively, with the presence of small rounded particles, which did not appear in BM. This phenomenon can be related to the abrasion and stirring action between the reinforcement particles and tool pin circumference, as well as to the extremely high plastic deformation occurred during FSW [9]. As a result of particle refinement, the average volume fraction of SiC in the matrix alloy increases to around 22%, 19%, and 21% in W1, W3, and W7, respectively. Fig. 5(a) shows the related EBSD Euler orientation and grain boundary maps (BM, W1, W3 and W7). It should be mentioned that for the cross-weld samples the locations of data acquisitions for EBSD analysis were focused on the centre of the NZ, to avoid the effect of complicated material flow in the top and bottom parts in this zone [34]. In these maps, individual grains were coloured according to their crystallographic orientations relative to the welding direction, and the black and red lines represent the HAGBs and LAGBs, where the white area represents the SiC. In terms of morphology, it can be seen that the evolved grain structures under different welding conditions were broadly similar to each other. In all cases, the grains were often not entirely delineated by a continuous HAGBs perimeter, and therefore the microstructure was represented in term of grain size and misorientation angles (a mixture of LAGBs and HAGBs) as shown in Fig. 5((b) and (c)). The grain size was measured based on either the area-weighted mean size D A or the arithmetical mean size D N [35]. BM microstructure shows a significant variation in the grain size ranging from 3 to 39 µm having a mean diameter of around 13 µm, with a mixture of HAGBs (high surface energy) and a high fraction of low energy substructure network LAGBs (subgrains). This can be related to the fact that when the polycrystalline metal deformed a large number of LAGBs are formed, as the LAGBs consist of arrays of edge dislocations and its structure and properties depend on the misorientation [36].
In comparison with BM, more homogeneous microstructures with fine equiaxed grains were formed in the NZ. At the lowest tool rotation speed of 1500 rpm, extra fine equiaxed grains having a size of around 2 µm were formed in the NZ combined by a low fraction of LAGBs ( Fig. 5-W1). Grain growth to about 8 µm happened with the increase of tool rotation speed to 2100 rpm ( Fig. 5-W7), as a result of high welding temperature, which is in agreement with a study conducted by Lee et al. [37]. In both cases, the grain boundaries are highly dominated by HAGBs. Consequently, the LAGBs increases with the increase in traverse speed, while the local misorientation change in a few grains, and it is more evident with 1500 rpm rotational speed (Fig. 5-W3). From the analysis of the misorientation angle distribution in Fig. 5(c)  The hardness profile exhibit asymmetric distribution with a typical "W" shape for all joints, similar to that seen in AA2009/SiC (AMC) FSW joints [38]. Regardless of the welding parameters, a softening process occurred in the FSZ, and the hardness value of the BM was much higher than that of the FSZ of around 150 HV. For the joints fabricated at the lowest traverse speed of 25 mm/min, a wider TMAZ and HAZ were formed with the increase of tool rotation speed from 1500 to 2100 rpm ( Fig. 6(a)), associated with a higher peak temperature and exposure time. However, the hardness value in the NZ was nearly the same in the ranging from 95 to 110 HV. AS the traverse speed increased to 100 mm/ min ( Fig. 6(b)), the average hardness of the NZ increased by about 20 HV combined by a reduction in the TMAZ and HAZ. Furthermore, the width of the high hardness zone in the joints fabricated at a traverse speed of 100 mm/min was larger than that of the joints welded at 25 mm/min. It is believed that this variation in the hardness of FSZ is a result of the difference of microstructures characterisations, which are controlled by welding parameters [37]. Fig. 7(a) shows the stress-strain curves of the BM, cross-weld (W1-W9), and longitudinal weld (WL3). Their ultimate tensile strength (UTS), yielding stress (YS), and percentage of elongation (% El) are compared in the bar graph ( Fig. 7(b)). Note that the cross-weld sample microstructure is heterogeneous as it is consist of four zones, i.e., BM, HAZ, TMAZ and NZ (Fig. 3(c)). Meanwhile, the longitudinal specimen contains only uniform fine equiaxed grains from the NZ. According to the stress-strain curves, the as-received AMC shows a tensile strength of about 419 MPa with a total elongation to failure of around 2.8%. In contrast, the cross-weld samples exhibit a higher ductility with a drop in the tensile strength as compared with BM. The reduction in tensile strength of AMC (AA6061/Al 2 O 3 ) FSW joints was also reported by Cavaliere et al. [18]. Furthermore, the stress-strain curve related to BM shows continuous yielding, while most of the welded samples exhibit a distinct yielding serration "jerky flow" phenomenon and similar findings have been reported by Mahmoudiniya et al. [39] and Pathak et al. [40]. Judging from Fig. 7(b), the mechanical properties of FSW joints are affected significantly by the welding parameters. It is shown that the joint fabricated at lowest heat input (W3) showing a higher strength (joint efficiency reaches about 75%) with the adequate elongation of around 5%. On the other hand, the joint welded at highest heat input and lower cooling rate (W7) shows a lowest joint strength (joint efficiency of about 52%) with a severe drop in the elongation. Surprisingly, the longitudinal welded samples (WL3) shows smooth curve characterised by a moderate combination between strength and elongation  of about 361 MPa (86% joint efficiency), and 8.5% (67% increase in elongation) than the BM, respectively. This can be related to the homogeneous fine grain structure in the entire sample. The location of the fracture in the typically tested samples is shown in Fig. 7((c), and (d)). Under most welding conditions, the FSW joints exhibited shear fracture mode occurred in the HAZ (AS) (see Fig. 7(c)) oriented at an angle of about 45°, as also presented by Moreira et al. [41], because of the loss of strength in this area as observed in the microhardness measurements. However, in W8 FSW joints fracture occurred in the stirring zone (see Fig. 7(d)), which can be related to macro defects on the top surface of the joint (Fig. 3(d)).

Low cycle fatigue life and cyclic deformation behaviour
The total strain amplitudes ∆ ( /2) εt as a function of the fatigue life (i.e., the number of cycle to failure, Nf) of the BM and FSW joints is plotted in Fig. 8. As it is evident, for a given total strain, the fatigue life of the FSW samples is lower than that of the BM. However, the results witnessed that, at high strain amplitude, the fatigue life show an apparent convergence in the BM and the welded joints. A similar trend has been also reported by Ceschini et al. [24] in FSW of AA6061 composite renforced with Al 2 O 3 . This behaviour can be attributed to the lower ductility of the composite metal. At high plastic strains amplitude, the cyclic ductility of the material limits the amount of induced plastic strain, and thus, the number of cycles that can be generated before fracture. The reduction in the fatigue life was more evident on the joints fabricated by using tool rotation speed of 1800 and 2100 rpm. On the other hand, significant enhancements of LCF property are achieved in the W1, and W2 even so at high strain amplitude (0.5%). This distinctive LCF performance is due to the fact that the formation of very fine equiaxed grains dominated by HAGBs in the NZ (see Fig. 5) combined with a high fraction of ultra-fine reinforcement particles can restrict the movement of dislocations most effectively [42]. Fig. 9((a) and (b)) illustrates the typical nominal stress-total strain hysteresis loops of the first and mid-life cycles at a given strain amplitude of 0.3% and 0.5% for the BM and W1. For the BM sample, the hysteresis loop area was nearly constant during the specimen life. Meanwhile, there is a significant change in the loop area in the case of FSW joints. These results are similar to those of FSW of AA7005/Al 2 O 3 [43]. By comparison of these figures, several features can be pointed out: i) For a given total strain, the amount of plastic strain is much higher for the FSW joints than for the BM, and this is in agreement with its lower fatigue life. ii) The evolution of the stress with the number of cycles in the BM was found to rise slightly, while in FSW joints, the cyclic hardening was more evident regardless of strain amplitudes. iii) Both of the BM and FSW joints exhibited nearly symmetrical hysteresis loops at both the outset and mid-life, representing an isotropic cyclic  hardening response. The extent of cyclic hardening is revealed by cyclic stress-strain curves and can be compared to the monotonic curves. Two approaches for generating cyclic stress-strain curves have been employed in this work. Saturated hysteresis loops and half of fatigue life cycle are used for the BM and FSW joints, respectively [44]. Fig. 9(c) compares the cyclic and monotonic stress-strain responses of the BM and typical FSW joint (W1). For both of them, it can be seen that the cyclic stress-strain curve lies above the monotonic stress-strain curve, indicating cyclic hardening behaviour.

Heat generation and microstructure evolution during FSW
Except for the plunging and retracting tool steps, heat is generated at a constant rate during the intermediate period, and similar geometry, structure and properties are formed in the cross-welds, which indicates a quasi-steady behaviour [45]. According to the concept of a pseudo heat index ω V ( / ) 2 , the heat generation in FSW is governed by two significant factors, i.e. the tool rotation and traverse speeds. The combined effect of these two factors is defined as the weld pitch (ratio of traverse speed to the rotational speed (mm/rev)), which is to control the energy generated per unit length [46]. Fig. 10(a) shows the effect of weld pitch on the maximum generated temperature in the NZ, indicates two tendencies for attention. Firstly, for joints welded with the same traverse speed, the reduction in peak temperature as a function of weld pitch was so sharp. For example, the increase of weld pitch by decreasing the tool rotation speed from 2100 to 1500 rpm (29%) caused the peak temperature to decrease by about 17%. Secondly, for joints welded with a similar rotational speed, a lower sharpness in the drop of welding temperature has obtained. The rises of traverse speed from 25 to 100 mm/min (300%) increases the weld pitch, lead to a reduction of the peak temperature by nearly 2.5,1.2% and 1.5% for tool rotation speed of 1500, 1800 and 2100 rpm, respectively. These findings are in agreement with the finding of Seidel and Reynolds [47]. They reported that the amount of energy input (Joules/meter) relates to the specific weld pitch, as it is increase, a reduction in the energy per unit length of weld occurred.
The observed microstructural difference between the BM and FSW joints are believed to be related to the temperature gradient, which is controlled by welding parameters [48,49]. Fig. 10(b) displays the relationship between the peak temperature and the mean grain size in the NZ. According to the figure, two behaviours are observed. Firstly, at the lowest tool rotation speed (1500 rpm), larger grains were formed by increasing the traverse speed (W3 > W2 > W1). This behaviour can be related to the incomplete recrystallisation in the NZ, as a result of lower heat input, insufficient plastic deformation, and high cooling rate, which is confirmed by the formation of low fraction of HAGBs (see Fig. 5-W3). Also, as the traverse speed decreases, the volume fraction of reinforcement particles increases due to SiC refinement (Fig. 4), therefore, the microstructure refinement is also attributed to grain boundary pinning with reinforcement particles [50]. Secondly, the results indicate structural coarsening by increasing rotational speed under the studied conditions, and this effect was found to be more pronounced at the highest rotation speed of 2100 rpm. In agreement with previous studies [12,34], the microstructure is coarsened with the increasing of welding temperature. On the other hand, finer recrystallised grain in the NZ was found accompanying the increase of the traverse speed, due to low exposure time at high temperature and high cooling rate as reported by Murugan and Kumar [17].
As a result, under all welding conditions, the grain structure in the NZ is characterised by smaller grain size and a higher fraction of HAGBs in comparison with BM due to continuous dynamic recrystallisation (CDR). However, the grain structure is different due to the complex interaction between the generated temperature and material flow during FSW. The mechanism of microstructure evolution continuously and uniformly dominated by HAGBs in the NZ can be explained by the hot working process under large strain deformation conditions. Firstly, a large number of dislocations were introduced in the stirring zone when the material deformed extensively and extruded through the confined area between the pin and the cold material. The dislocation climb, cross-slip, and glide occur quickly and extensively at high temperatures in metals of high stacking fault energy (SFE) such as aluminium (around 170 mJ m -2 ), which is the mechanism of dynamic recovery (DR), and results in the formation of more organised dislocations arrangement (LAGBs) (subgrain). The subgrain size can be related to Zenner-Hollomon parameter (Z), as the material plastic flow at an elevated temperature where the deformation and softening processes occur depends on the deformation temperature, strain rate, and strain [51]. Recently, Mahmoudiniya et al. [39] reported an empirical equation for predicting the final grain size of FSW DP700 steel based on the Z parameter. They mention that the increase in temperature decreases the Z parameter and increases the recrystallised grain size. This gives an explanation to the coarsening of the microstructure with increasing rotational speed because the Z parameter experienced a decreasing trend with increasing rotational speed (higher welding temperature). Secondly, the CDR process occurs as the increasing of applied strain resulting in introduced dislocations continuously to the subgrains by further deformation. As the subgrain size is almost independent of the strain during deformation at high temperature, instead the subgrains rotate as they accommodate more dislocations into their boundaries, forming equiaxed grains having a size comparable to the subgrain with HAGBs [52][53][54]. This gives an explanation to the dominating of HAGBs on the microstructure of the NZ. Both of the DR and CDR processes are essential in FSW, as they lower the flow stress of the material, so enabling it to be deformed more efficiently. The effect of FSW parameters on the hardness profile of AMCs joints displayed in Fig. 6. According to the figure, the hardness profile of the FSZ is dependent on the process parameters, and the measured hardness value is lower than the BM. As reported in the other studies, the reduction in hardness is usual in heat-treatable artificially aged aluminium alloy (6xxx-T6 series) [28,54,55]. Microstructural evolution due to thermomechanical effect during FSW, in term of strengthening precipitate, and grain size can be considered as two mechanisms for softening in FSW joints. The hardness value decreases in the HAZ, although there is no apparent difference in the microstructure from that of the BM, indicating that the thermally induced cycle during FSW has a significant influence on this zone. Also, the hardness reaches a minimum in the TMAZ region, as it is exposed to the grain deformation and thermal cycle. As there is no change in the size or shape of reinforcement particles in the HAZ and TMAZ, the softening process can be attributed to the annealing or overaging of strengthening precipitate. According to Gatea et al. [56], the hardness of AA 6092/SiC/17.5p dropped when heat treated to 415°C (O-condition) as compare to T6condition, which was nearly the same as the temperature in the HAZ and TMAZ of FSW joints. They claim this is due to the reduction of Si concentration in the matrix/particle interface and dissolution of Al 2 Cu (orthorhombic crystal structure) and reprecipitated Al 4 Cu 2 Mg 8 Si 7 (hexagonal crystal structure). Also, Olea et al. [57] claim the reduction in the hardness of the TMAZ to the overaging of strengthening precipitate, as transformed from needles to lath morphology.
The most significant recovery in hardness is observed in the NZ, which can be attributed to the refinement of both the matrix grain and SiC particles and the generation of dislocations at the matrix-SiC interface due to high differences in the thermal expansion coefficient between SiC and aluminium matrix [9,16]. Furthermore, the NZ experiences the highest temperatures generated by the welding process (above the solvus of the initial precipitate of around 540°C), leading to the dissolution of precipitates and the formation of a supersaturated solid solution (SSSS). The SSSS undergos expected natural ageing and hence increases the hardness [58]. Fig. 11 presents the relationship between the tensile strength, fatigue life and mean grain size. According to Fig. 11(a), two opposite tendencies can be observed relating the effect of grain size on the tensile strength. Firstly, at the lowest tool rotation speed of 1500 rpm, the tensile strength of FSW joints is improved with an increase of mean grain size (W3 > W2 > W1). The increase in monotonic strength of FSW joints can be attributed to several mechanisms. Chawla et al. [5] proved that the strengthening of AMC is significantly affected by the microstructure of the aluminium matrix and reinforcement particles. For the studied BM in T6 metallurgical conditions (strengthened by strengthening precipitates), a higher traverse speed results in insufficient stirring time for complete recrystallisation (producing a coarse grain structure). This, combined with the partial dissolution of the precipitates resulting from the FSW thermal cycles and subsequent re-precipitation during natural ageing after FSW, results in an increase in strength [32].

Structural property in the friction stir welded AA6092/SiC
Furthermore, according to the Taylor strengthening mechanism, volumetric strain occurs in the composite when using a fast cooling rate due to the thermal mismatch between the aluminium matrix and SiC particles which leads to the generation of geometrically necessary dislocations (GNDs) around the reinforcement particles. As the interaction of the mobile dislocations increases, flow stress of composite metal increase [59,60]. This mechanism is confirmed by the formation of a high fraction of LAGBs substructure at higher traverse speed (Fig. 5-W3).
Secondly, at higher tool rotation speed (1800 and 2100 rpm) the tensile strength of weldments goes through a maximum with a higher traverse speed as a result of the evolution of entirely recrystallised finer grain. This is in agreement with the Hall-Petch strengthening mechanism, as grain refinement and the formation of a high fraction of HAGBs leads to the increase of the resistance to the motion of dislocation and pile-up of dislocations at grain boundaries, which plays a significant role in enhancing resistance to plastic flow [61]. The effect of grain boundary on the strength of metal is also confirmed in Al-7% Si cast alloy and SAE 1020 steel [36,62].
The fatigue properties are frequently correlated with tensile properties [36]. In LCF the strain resistance depends on the ductility of the metal, i.e. higher ductility higher the fatigue life. Although the tensile characterisation of the welded joints showed higher ductility than the BM, the fatigue life of FSW joints is lower than the BM and vary based on the welding conditions (see Fig. 8). The variation in fatigue behaviour is mainly dependent on the evolved microstructure characterisation and cyclic strain range [63,64]. Fig. 11(b) displays the relationship between the fatigue life and grain size at different strain amplitude. As shown in the figure, two opposite tendencies are pointed; firstly, at lowest tool rotation speed (1500 rpm), the increase of traverse speed leads to lower fatigue life (W1 > W2 > W3) under all strain amplitudes, this trend is opposite to that of tensile behaviour ( Fig. 11(a)). Secondly, at high tool rotation speeds of 1800 and 2100 rpm, faster traverse speed improves the fatigue life of FSW weldment, i.e. W9 > W8 > W7.
For both trends, it can be seen that the fatigue life has been improved with the formation of smaller grain size combined with the high fraction of HAGBs (see Fig. 5), and high volume fraction of SiC. The formation of HAGBs almost leads to high ductility, and as the diameter of the grain is reduced more of the grain boundaries properties will be reflected in the centre of the grain. Therefore, this can effectively prevent macroscopic plastic deformation, and minimise microscopic localised plastic strain under specific cyclic stress amplitude. Also, in AMC, an increase in the volume fraction of the hard ceramic reinforcing particulates (due to the particle refinement) increases dislocation density and reduces grain size. These microstructural changes tend to increase the resistance offered to the motion of mobile dislocations. Furthermore, load transfer from the matrix to the reinforcement is a significant contributor to composite strengthening [59,65]. The cyclic hardening behaviour of the BM and FSW joints during fully reversed cyclic straining under all strain amplitude (Fig. 9) could be attributed to the following reasons; (i) The competing and synergistic effects of stress (load) transfer between the soft and ductile aluminium alloy matrix and the hard and brittle ceramic reinforcement particle (SiC) [66]. (ii) Grain refinement, as the grain boundaries act as barriers to dislocations movement, thus the strain hardening in the smaller grain size is higher than the coarse grains [67]. (iii) The preexistence of high dislocation density in the aluminium matrix produced by the presence of the SiC reinforcement due to a mismatch in the coefficients of thermal expansion [68,69]. (iv) An increase in the storage dislocation density during cyclic plastic deformation, as they accumulate by randomly trapping each other, which is associated with plastic strain, or they are required for compatible deformation of various parts of the crystal (GNDs) associated with plastic strain gradient [64].

Conclusions
In the current work, the effect of FSW parameters (tool rotation and traverse speed) on the heat generation and microstructure evolution of AA6092/SiC/17.5p-T6 joints and its effect on the tensile and LCF properties are systematically examined. The rotational speed has a more significant effect on the generated peak temperature, while the traverse speed controls the exposure time and subsequent cooling rate. The microstructure of NZ exhibits an elliptical shape, and fine equiaxed grains resulted from the CDR process. Grain growth in the NZ occurs as a result of incomplete CDR or in particular when exposure time at high temperature is very long. The grain boundary in the NZ is a mixture of LAGBs and HAGBs, and it is fraction controlled by the traverse speed. The SEM micrographs reveal a homogeneous distribution of reinforcement particles (SiC) in the NZ with an increase in the volume fraction because of particles refinement. The microhardness distribution of the welded joints under different welding conditions exhibited typical ''W'' shape features. The width of the minimum hardness zone varied with the induced temperature and corresponded accordingly with the range of the HAZ and TMAZ. The tensile properties of the joint welded at a rotational speed of 1500 rpm and traverse speed of 100 mm/min are the best amongst all weld joints tested. The joint efficiency of the welded joint reached 75% of the BM, while the elongation increases by 44%. Increases in LCF life of the AA6092/SiC/17.5p FSW joints were observed with decreased grain and reinforcement particle size and increased particle volume fraction. FSW parameters control these microstructural features, and it is found that the combination of tool rotation speed of 1500 rpm with a traverse speed of 25, and 50 mm/min resulted in optimal joints performance. The cyclic stress-strain curves of the FSW composite showed evidence of progressive isotropic hardening to failure for all cyclic strain amplitudes, while the BM showed less hardening ability.