An experimental study on lap joining of multiple sheets of aluminium alloy (AA 5754) using friction stir spot welding

Friction stir spot welding (FSSW) process is widely used in the automotive industry for a range of applications such as battery components, standard wire connectors and terminals. This manuscript addresses two grand challenges in the arena of FSSW, hitherto, unaddressed in the extant literature: (i) lap joining of thin sheets (0.3 mm thickness) of AA 5754 alloy and (ii) lap joining of more than two sheets using FSSW. To accomplish this task, a novel pinless convex-shaped tool was designed to alter the stress state while gradually advancing the tool which led to achieving stress state necessary for obtaining defect-free lap joints. The weld joints were inspected by optical microscopy, SEM imaging and analysed by nanoindentation tests and Vickers microindentation tests for assessment of the quality of the weld interface (WI). Process parameters of FSSW such as torque on the tool and axially applied load were used to analytically obtain the average local measure of peak normal and axial stresses as well as the coefficient of friction in the contact zone. In samples welded at low rotational speeds, the strain-hardening mechanism was seen dominating in contrast to samples welded at higher rotational speeds, which showed thermal softening. As a direct consequence of this, the samples welded at low rotational speeds showed much higher hardness at the weld surface than the samples welded at higher speeds. A strong transition of strain hardening to thermal softening was noticeable beyond an applied strain rate of 400 s−1.


Introduction
The substitution of copper (8.96 g/cm 3 ) by aluminium (2.7 g/ cm 3 ) is being sought as an immediate engineering requirement for battery components, strand-terminal connectors and terminals to save weight and transportation costs. However, an issue that appears in the literature is the reliability of joining of aluminium alloy sheets. Therefore, the development of joining technologies for aluminium alloys with acceptable weld characteristics and minimal energy consumption is emerging rapidly [1][2][3].
Currently, the most frequently used spot joining techniques for assembling aluminium alloy sheets are resistance spot welding (RSW) [4], cold metal transfer spot welding [5], laser spot welding [6], ultrasonic spot welding (USW) [7], selfpiercing riveting [8] and friction stir spot welding (FSSW) [2]. In contrast to other technologies, FSSW consumes low energy, and because of being a solid-state welding process, it can provide superior mechanical properties to the processed material compared to the base material. Hence, there lies a considerable incentive for research on this topic.
Friction stir welding (FSW) is the technique which was developed at The Welding Institute (TWI) of the UK in 1991 [9]. Initially, it was applied only to welding, but over the years, the technique has gained popularity for use in surface property enhancement, channelling, riveting, forming and many more other processes. A series of developments in the field have led to the emergence of latitude of processes that are all based on harnessing the principle of friction stir processing. Fig. 1 shows an attempt to classify and summarise these friction-based manufacturing processes by consolidating the literature available on this topic.
Conventional FSSW is a solid-state welding technique that was developed by the German research centre GKSS. Mazda was among the first companies to report the first application of FSSW for lap joining during mass production of its 2003 RX-8 cars [2,10].
FSSW technique relies on plunging a rotational tool (consisting of a pin and a shoulder) into the overlapping area of the two sheets until the shoulder of the tool touches the surface of the upper sheet. Thereafter, the tool is retracted and leaves a keyhole behind [11][12][13][14]. The keyhole represents the biggest drawback of the conventional FSSW process. It has been shown that the keyhole reduces the effective cross section area of the joint, which adversely affects the mechanical properties of the joint [15] and this becomes the potential site for corrosion [16,17]. Two methods were proposed to fill the keyhole, refill friction stir welding (RFSSW) [18][19][20][21] and intermediate layer friction stir welding (IL-FSW) [22], but both RFSSW and IL-FSW demand complicated equipment. The most common problems known to occur in all types of FSSW joints are hook defects (HD). It is shown by the researchers that HDs occur during IL-FSSW [23] and RFSSW [24] and they tend to reduce the weld joint strength.
Bakavos et al. [25] studied FSSW of 0.93-mm-thick aluminium copper alloy, with a flat pinless tool (Ø 10 mm) with five different surface features and cycle (welding) time varying from 0.5 to 2.5 s. High lap shear strength (~3.4 kN) was achieved; however, different types of defects like HDs, flashes, circumferential shear crack and unwelded and/or partially bonded weld joints were observed. Pinless FSSW offers a better alternative to overcome the aforementioned problems as the joining during this process is assisted by dynamic recrystallisation and diffusion bonding between the two layers [26]. Both mechanisms occur above the temperature of recrystallisation. However, diffusion prevails under the influence of high axial stress while dynamic recrystallisation is driven by high shear stresses in the contact zone. State-ofthe-art concerning the topic of FSSW suggests that despite some success, it has not been possible to use FSSW to fabricate sheets of thickness less than 0.93 mm. Moreover, the literature on this topic lacks evidence on whether more than two sheets were welded at a time, so it has been assumed here that joining multiple thin sheets is an ongoing challenge. After identifying the knowledge gaps discussed above, several key questions emerged that will be addressed as part of this research effort: (a) Can the pinless FSSW method be used to lap join multiple sheets of thickness as low as 0.3 mm? (b) If yes, what should be an ideal tool geometry of the pinless tool to achieve this task?
(c) What is the influence of the rotational speed of the tool on the microstructure of the weld shape and properties of the weld joint? (d) Whether the pinless FSSW process is dominated by strain hardening or thermal softening and what magnitude of strain rate governs these two mechanisms?
The above questions motivated the design and development of a new tool in this work with a geometry capable of welding multiple thin sheets of a softer aluminium alloy. The overarching motivation of this paper stems from the fundamental curiosity of studying the allied process characteristics of the pinless FSSW process by complete characterisation of the mechanical properties to achieve a defect-free weld joint.  Fig. 1 Classification of the friction stir-assisted manufacturing processes alloy 5754-H111) that was measured using an optical emission spectrometer (ARL 3580). Before to welding, the base material was mechanically tested for its uniaxial stress-strain property using a Hegewald & Peschke Inspect Retrofit universal testing machine. The tensile stress-strain curve obtained for the base material is shown in Fig. 2, and in light of the experimental results discussed in the latter part of the paper, this plot was used to draw a sound comparison with the analytical stress analysis obtained from the data.  Fig. 3b shows the tool geometry that was used to lap join four thin sheets during this work. The tool was produced in the lab using H13 (X40CrMoV5-1) hot-work tool steel with its nominal chemical composition shown in Table 2.

Working principle
The welding system consists of a clamping system (used to constrain the material during FSSW) movable in the vertical direction, providing sample clamping as shown in Fig. 3a. The way this machine works is that the sample is firstly held by the clamping system, and during this time, the preloaded tool starts to come vertically downwards onto the upper sheet. The tool rotation initiates as soon as the system detects the tool touch onto the surface of the upper sheet. The time in which the tool completes its operation (cycle time) depends on the resistance of the material (soft materials are easily displaced), applied axial load, rotational speed and the total tool travel distance. Upon reaching its target vertical displacement, the tool retracts quickly. During the FSSW process, a Ktype thermocouple (Fig. 3) was placed at the centre of the bottom of the substrate with a view to gaining information on the variation in temperature during the different runs carried out at different rotational speeds.

Proposed convex tool design considerations
After a series of dummy trials, it was found that the traditional flat pinless tool did not work for joining multiple thin sheets. As opposed to a flat tool which causes tearing of sheets (highlighted in red circles in Fig. 4) due to a large contact radius, a convex tool presents a gradually increasing contact To produce a high-strength multilayer welded joints using the pinless tool, it is necessary to produce a certain level of stress to arouse material flow, that is, to generate conditions for dynamic recrystallisation as well as diffusion between the two interfaces [9]. During FSSW, the material flow is most pronounced at the maximal contact radius of the tool where the angular velocity of the tool is maximum. However, at the centre of the tool, the angular velocity is theoretically zero. In this small area, the material is compressed due to the axial pressure present in the contact zone causing an individual grain to yield and with increased penetration, resulting in dislodging of the adjacent grain. In a nutshell, this means that a flat tool creates a uniform pressure along the entire radius of

Process parameters
FSSW experiments were carried out on an adapted force-controlled EJOWELD C50R FSSW machine which is designed to run at the rotational speed of up to 9000 rpm, maximum axial load of 8 kN and the maximum cycle time of 5 s. To start the process, the tool was made to contact the surface of the top sheet to be welded and then it was rotated along the axis normal to the sheet surface at a fixed RPM. A predefined constant load was applied to achieve the desired contact depth (usually less than the thickness of the upper sheet).
During the experiments, the tool experienced an upward normal force termed as an axial load which was maintained at 2 kN throughout the experiments reported in this work. The axial displacement made by the tool at this load (often referred to as 'plunge depth') was maintained at 0.25 mm in all the experiments. During the trials, rotational speed was varied in increments of 500 rpm ranging from 1500 to 3500 rpm as shown in Table 3.
The focus of experiments was to collect the experimental data at five rotational speeds of 1500, 2000, 2500, 3000 and 3500 rpm. Each experiment was repeated three times and an average reading was used to report the results in this work.

Material characterisation of the welded joints
After FSSW, the metallographic examinations were performed. As for this, the welded joints were ground with sandpaper (grit 360 to 2500) and polished with diamond suspensions (6 and 3 μm). The final polishing step was done with 0.05-μm colloidal silica. Electrolytic etching was done on Struers LectroPol-5 device (40 V, 2 min, Barkers etchant). The cross-sectional morphology of the joints was analysed by a light microscope (Zeiss AxioScope.A1 with AxioCam ICc3) for etched samples and a Zeiss Axio Vert.A1 MAT with AxioCam 105 with crossed polarised light and sensitive tint was used to analyse polished samples. Further characterisation of samples was carried out using the field emission scanning electron microscope (FE-SEM) (Hitachi S-4800).

Microhardness mapping using Vickers tests
As for the evaluation of the local variation in mechanical properties, microhardness imaging of the weld joints was obtained using standard Vickers microhardness tests (Struers DuraScan 70 machine) at a load of 0.1 kg. Fig. 5 shows an  Fig. 7 a-e Evolution of applied axial stress and shear stress over time estimated from the applied load and torque data at different RPMs starting from 1500 to 3500 rpm. The plots are also compared against the yield limits and tensile strength of Al 5754 base metal obtained from the tensile stress-strain data of AA 5754 b illustrative scheme that was implemented to setup the machine to obtain the cross-section hardness imaging of the welded specimens. The interspacing of 0.33 mm between the indented points ensured that there was little or no reflection of stress waves induced in the material during each indent cycle.

In situ scanning probe microscopy (in situ SPM) using Hysitron Nanoindentation
Beside Vickers test, a more refined nanomechanical hardness testing was made using scanning probe microscopy (SPM) with an instrument from Bruker Hysitron XPM (TI 980 Nanoindenter) equipped with a Berkovich tip. The XPM load function was set to be 20 × 20 indents, 500 nm separation and 150 μN peak load. Almost about 400 indents were performed in about 3 min. A sample result shown in Fig. 6a marked with a red square (in the joint interface) highlights a small area (25 μm × 25 μm) mapped by hardness testing while the welding was performed at 1500 rpm. A fully automated testing was performed on both sides of the weld faying interface (WFI) and measurements were stitched to obtain a mapped image shown in Fig. 6b.

Stress analysis, hardness mapping and microstructural variation in the welded zone
The apparatus used to perform FSSW was equipped to provide the measurements of torque (N·m) and cycle (welding) time (s), while the temperature (°C) information at the bottom of the sample was obtained via thermocouple as explained earlier via Fig. 3. Post-processing of the data was carried out by using analytical formulae shown below to estimate the coefficient of friction, axial stress and shear stress in the contact [27,28].
where R is the fixed tool radius (mm) and r(t) is the instantaneous contact radius r t (mm) calculated from the data of axial displacement of the tool (plunge depth) (mm) h(t), F(t) is the applied axial load measured during the process (N), T(t) is the torque (N·mm). σ n (t) and τ(t) refers to axial and shear stresses (MPa), respectively, and μ (t) is the coefficient of friction. At the beginning of the operation, the axial load was set to be 2 kN. Fig. 7 shows the evolution of two major stresses σ n (t) and τ(t) in the welded zone, namely the axial stress presented by the virtue of applied axial load and the shear stress estimated from the torque data presented by the virtue of tool rotation. During early stages of the contact, it is primarily the axial stress that helps the tool to plunge into the workpiece surface and this gives rise to the stick type of contact conditions between the tool and the workpiece.
As Fig. 7 shows, the increasing penetration of the tool vertically downward (specifically for the convex tool) causes the magnitude of shear stress to rise and it scales linearly (seen by a sudden sharp increase). At higher speeds, slip dominates sticking and the competition between these two stress states influences the residual stress state on the welded surface. Moreover, a steady-state welding condition was obtained after the competing stick-slip conditions stabilised the toolworkpiece contact.
Particularly noteworthy from the stress profiles is that during the early stages of contact, the axially applied stress dominates the shear stress as the area of contact between the tool and workpiece is relatively smaller. It means that before a certain threshold area activates, the contact between the tool and workpiece did not achieve a steady-state for the plastic flow of the material, particularly until the slip (shear stress) does not start to dominate over the stick (axial stress). Reilly et al. [29] simulated the material flow phenomena during FSSW process. A close inspection of their simulations and the experiments performed in this work indicates that during the initial stages of the contact, the centripetal velocity drags the material inwards underneath the tool. It can analogously be understood or seen by visualising the pathline of a particle at an imaginary radius (outer periphery) 'x' such that the particle at the peripheral radius of 'x' is pulled in a spiral net inward and underneath the tool. This portion of material trapped between the tool and workpiece continues to churn under the rotational action of the tool with increasing downward displacement of the tool. The shear stress during this time tries to overcome the work of displacing this portion of material from radius 'x' to inside of the core. As this portion of the material deforms, the shear stress goes up before eventually reducing down. This point signifies the transition from unsteady tool-workpiece contact to a steady contact such that the shear stress beyond this point decreases almost monotonically as evident from Fig. 7. Fig. 8 shows the variation in the coefficient of friction at five rotational speeds of FSSW examined in this work. The value of CoF varies between 1 and 2.5 depending on the speed of operation. A higher value of CoF at the end of the process was observed to be responsible for strain hardening of the surface (e.g. 1500 rpm) where CoF was about 1.75 at the end of operation as opposed to the lower value of CoF of 1.14 at the start of the process. Farhat et al. [30] and Kumar et al. [31] reported that the FSW process at low rotational speeds (from 200 to 1400 rpm) is governed by strain hardening.
The phenomena of strain hardening at low RPMs was also verified by examining the Vickers hardness imaging (shown in Fig. 9) revealing that the microhardness on the surface was highest in the welded zone for samples welded at low speeds. Overall, in samples welded at 1500 rpm and 2000 rpm, the strain hardening mechanism dominates over thermal softening, and conversely, in samples welded at 3000 and 3500 rpm, thermal softening was more pronounced.
To augment support to the argument about strain hardening mechanism revealed by Vickers tests, nanoindentation was used for further analysis of the strain-hardened region. Fig. 10 in  Table 4 shows three different clusters (marked in red, green and blue) in strain hardened region revealing variations in hardness and elastic modulus probed by the nanoindentation mapping. From a quantitative analysis, the first and second cluster seems to correspond to strain hardened material in the stir zone, while the third cluster (appears discretely and barely very few in numbers), showing highest values appears to be intermetallic brittle phase of AA 5754 alloy.
It is noteworthy that the samples welded at 1500 and 2000 rpm showed a serrated stress-time curve (jerky flow) seen earlier in Fig. 7. According to Yamada et al. [32], the appearance of jerky flow signifies that the solute Mg atoms as well as precipitates in Al-Mg alloy interact with dislocations. This phenomenon is known as Portevin-Le Chatelier (PLC) effect. Materials like Al-Mg alloys possess the tendency to exhibit PLC effect while they undergo inhomogeneous plastic deformation. Furthermore, under conditions of local increase of the strain rate above certain limit leads to softening of the material locally. Next, the formula proposed by Chang et al. [33] was used to perform calculations for estimating the applied strain rate as a function of rotational speed.
where r e and L c are the effective (average) radius and depth of the stir zone (recrystallised zone) measured with the help of ImageJ software from the cross-sections of etched samples. Chang et al. [33] estimated average material flow Rm to be half of the tool rotation speed (Fig. 11). Gerlich et al. [34] estimated strain rate at different rotational speeds (from 750 to 3000 rpm) in Al 5754 CFSSW samples by using Zener-Holloman parameter and obtained strain rate vs. rotational speed curve close to the curves presented in this paper. It can be seen that during FSSW, the strain rate increases with increasing rotational speed. According to stresscycle time plots, CoF vs. time plots and hardness measurements, it can be assumed that strain hardening is more pronounced at strain rates lower than 400 s −1 and thermal softening is more pronounced at strain rates > 400 s −1 .
To lend credence to this observation, the grain size conditions were assessed by subjecting the cross-section of the welded zones to electrolytic etching. The variations in the grain size are shown in Figs. 12 and 13, respectively. It was observed that the low RPM (low strain rate < 400 s −1 ) leads to small grain sizes while high RPM (strain rate > 400 s −1 ) leads to coarser grains. Jata et al. [35] proposed continuous dynamic recrystallisation (CDRX) as grain refinement mechanism during FSSW. This process is driven by temperature, strain and strain rate as the well as interaction between precipitates and solute atoms with dislocations. In a few references [21,[35][36][37], it was revealed that the magnitude of misorientation angle increases significantly during FSSW/FSW compared to the base material. Thus, new grains seen in the stir zone can be disorientated subgrains. Repeated dislocation absorption into subgrain boundaries is the dominating mechanism during CDRX for increasing the misorientation angle between adjacent subgrains [9]. Therefore, it is proposed that the microstructure in samples welded at different rotational speeds is  . 11 Relation between tool rotational speed and strain rate influenced by (i) grain size reduction by subgrain rotation induced by dislocation glide and (ii) subgrain coarsening where average subgrain size increases while a number of subgrain decreases and this is driven by the reduction of local misorientation associated with intracrystalline recovery. The former is more dominating in samples where strain hardening overcomes thermal softening (lower rotational speeds) while the latter is more dominating in samples where thermal softening is more pronounced.

Characteristics of the weld joint shape
Fig. 14 shows an outcome of the entire FSSW process revealing detailed nature and shape of the welded joint interface. Of importance to note from Fig. 14 is the presence of edge chipping on the outer periphery of the contact radius. The examination of this cross-section reveals a strong presence of fracture as opposed to plastic deformation (white arrows) in Fig. 12 at 3500 rpm. It is clear that the sheets on the outer side of clamping tend to go upward during FSSW but are constrained by the clamping system. Consequently, there are opposite thermal stresses developed radially (σ r ) in the sheet. Due to continued downward movement of the tool, this radial stress grows further and results in material going upwards (although stochastically) to come in contact with the tool. The portion of sheet material that comes in contact with the tool is smeared away and leaves behind the sites of fracture. Earlier, it was shown that the edge chipping [38] contours are rather complex and shos no regularity. Also, the parameter "fractal mean square deviation" is quite useful in the evaluation of this contour profile. What is apparently clear though is that when the magnitude of tensile stress goes past a certain threshold value exceeding the fracture toughness of the metal, it causes chipping rather than plastic deformation as evident in this work. Fig. 15 shows variations in several physical parameters measured as a function of the rotational speed of the tool. Those parameters include for instance the measurement of d  After careful data analysis, it became clear that the actual advancement of FSSW tool into the workpiece during the process appears to be consistently higher than the programmed depth of advancement (shown in Fig. 15 as inverted from the bottom). This could be attributed either to the positioning errors or to material induced plastic effects in the stir zone (contributing to different viscosity and permeability of material for the tool to go past) that impedes the sensor from precisely detecting the surface of the sheet, which was being considered as the theoretical datum for measurement reference.
Most notably, it was observed that an increase in the rotational tool speed led to an increase in the D parameter. It simply means that a higher RPM produces samples with a conical shape and taper angle 'α'. The taper angle 'α' decreases linearly with a decrease of the rotational speed of the  The parameter d/D ratio was observed to decrease with an increase of rotational speed. The ratio of 1 would make the shape to be cylindrical. During steady state weld, the axial load applied on the workpiece pushes the softened material up and lacks sufficient stress in the lower part of the sample, which causes D to increase while d and H decreases. As rotational speed increases, the cycle time decreases but the temperature increases (see Fig. 16) causing thermal softening of the material.
Usually, during FSSW, the presence of hook defects, wormhole or partially bonded areas are presented as major type of defects in the weld joints [39][40][41]. However, the scarcity of literature on FSSW of sheets of thickness less than 0.93 mm does not confirm this and it is possible that the hook defects appear only above some critical thickness.
To analyse this, SEM inspection of the weld interface was made as shown in Fig. 17. Fig. 17 corresponds to the same locations as shown in Fig. 12a with white and red squares for bonded regions between the first (uppermost) and second sheets and between the third and fourth, respectively. It was observed that (a) a fully bonded region was obtained which did not show any obvious boundary at the interfacing surfaces, whereas in some other regions, as shown in Fig. 17b showed few partially bonded regions characterised by intermittent welded areas and intermittently present interfacial boundaries were noticed especially pronounced at higher rotational speeds. The samples welded at 3500 rpm were found to have occasional wormhole defect near to the centre of the sample and the last two sheets were observed to be only partially bonded.

Conclusions
FSSW has remained one of the thrust areas in joining technology of manufacturing over the past few years. The two grand challenges identified from the extant literature addressed as part of this research are (i) FSSW of 0.3-mm-thick sheets and (ii) FSSW of more than two thin sheets. By designing a novel pinless convex-shaped tool, the paper reports achieving successful defect-less lap joining of a number of aluminiummagnesium (5754 H-111) sheets. During the course of this research, a couple of new scientific insights were gained about the FSSW process and following broad conclusions were obtained:  i) Whilst using a convex-shaped pinless tool, there is a trade-off for choosing the rotational speed responsible for the alteration of the microstructure. It was found that the applied strain rate (rotation speed of the tool) has a direct influence on the microstructure of the joined part, i.e. low rotational speed leads to smaller grain sizes (grain refinement causing higher microhardness of stir zone) and a higher rotation speed typically leading to an applied strain rate > 400 s −1 leads to thermal softening. (ii) The increase in microhardness of the FSSW was found to be accompanied by the dominating effect of strain hardening, due to the competing mechanism of shear stress and axial stress present in the weld zones. This competition leads to cthe occurence of stick-slip until the onset of steady-state weld conditions. (iii) An increase in the rotational speed changes the d/D ratio causing a change of the weld pool profile from being cylindrical to conical. Also, it decreases the thickness of the welds in the centre of the stir zone. (iv) Until a certain threshold rotational speed, the quality of the welded joint is better and did not show any defects and partially bonded regions, but higher speed arouses the possibility of developing defects in the stir zone. Weld interfaces showed uniform mechanical properties (hardness and modulus) regardless of the rotation speeds.
(v) Overall, the variation in the temperature in the FSSW process was seen to be relatively low when compared to the variation in the shear stress during the process.
Consequently, it appears that the FSSW process is primarily driven by the presence and domination of the ratio of shear and axial stress making the FSSW process to be stress dominated rather than temperature dominated. Least to mention that the ratio of the two stresses directly refers to the coefficient of friction and thus the name of the process being friction stir welding is fully justified in this context.