G Model MSA-25619; No. of Pages 7 ARTICLE IN PRESS Materials Science and Engineering A xxx (2009) xxx–xxx Contents lists available at ScienceDirect Materials Science and Engineering A journal homepage: www.elsevier.com/locate/msea Effects of thermal boundary conditions in friction stir welded AA7050-T7 sheets P. Upadhyay, A.P. Reynolds ∗ Department of Mechanical Engineering, University of South Carolina, 300 Main Street, Columbia, SC 29208, USA a r t i c l e i n f o Article history: Received 30 July 2009 Received in revised form 12 October 2009 Accepted 20 October 2009 Available online xxx Keywords: 7XXX series alloys Boundary condition Under-water FSW a b s t r a c t A series of friction stir welds was made in laboratory air and with the plates submerged in water to investigate how the quenching rate affects properties of the joint and some weld response parameters. Select welds were also made at a sub-ambient temperature of −25 ◦ C. Temperature measurements were made in the probe center and at the minimum hardness location of the weld. Weld response variables, hardness distributions, joint strength and nugget grain size were measured and correlated with boundary conditions and welding parameters. A consistent decrease in the peak temperature and increase in cooling rate were observed in the submerged welds. Submerged welds show improvement in tensile strength and elongation throughout the range of parameters tested. © 2009 Elsevier B.V. All rights reserved. 1. Background and introduction Peak temperature and the rate of cooling during the friction stir welding (FSW) process are key parameters that dictate the weld properties. It has been reported in the literature [1,2] that all other things being equal, the peak temperature in the stir zone is proportional to the tool rotational rate, whereas the cooling rate and, hence, length of time of stay above a certain temperature is dependent on welding speed. Sato et al. [1] have reported the effect of welding parameters on the peak temperature in the stirred zone for AA6063. The study reported that the peak temperature increases with the increase in tool rotational rate at constant welding speed, the rate of increase in temperature being less with high rpm. Temperature simulation by Reynolds et al. [2] has shown that transient length of thermal history is governed by the welding speed. Higher the welding speed shorter will be the time for which the stir zone stays above an elevated temperature. In other words, for welds performed under standard, ambient conditions, the heating and cooling rate is dependent on the welding speed. It is intuitive and has been demonstrated [3], that the heat extraction rate can also be increased by employing rapid cooling techniques such as, welding under water and in the presence of cold fluids. Welds have been performed under water in offshore structural repair and development using conventional welding methods like shielded metal arc welding for a long time. These methods suffer from issues like hydrogen embrittlement, oxidation and porosity which worsen at greater depth [4]. Being a solid state joining pro- ∗ Corresponding author at: Department of Mechanical Engineering, University of South Carolina, 300 Main Street, Room A224, Columbia, SC 29208, USA. Tel.: +1 803 777 9548; fax: +1 803 777 0106. E-mail address: reynolds@cec.sc.edu (A.P. Reynolds). cess friction stir welding bypasses these problems. Some references can be found in the literature concerning friction stir welding performed submerged in water and other cold fluids and property alterations associated with it [3,5–9]. Early work in submerged friction stir welding performed at low ambient temperature can be attributed to Benavides et al. [5]. They reported a significant reduction in stir zone temperature when AA2024 plates were friction stir welded submerged under liquid nitrogen. Although the joint suffered from worm-hole defect, a significant reduction in stir zone grain size to 0.8 ␮m was reported [5]. Sakurada et al. used the inertia friction welding method to join AA6061 under water [6]. Comparing the weld with regular in-air welds, they reported an increase in joint strength and decrease in the width of heat affected zone. Nelson et al. [3] studied the effect of quench rate on 7075 and 2195 aluminum alloys. They performed friction stir welding by externally heating and cooling the parent metal plate and anvil. Cold water and mist were used to chill the plates in the wake of the tool. They reported a maximum of 10% increase in tensile strength with respect to conventional FSW of AA7075 after post-weld natural aging of 1000 h. Staron et al. applied liquid CO2 coolant near the weld seam to investigate residual stress improvements [7]. They report significant reduction in tensile residual stress for 6.35 mm thick AA2024 plate. Su et al. reported production of nano-scale grain size in 7075 Al sheet by quenching the plate behind the tool with a mixture of water, methanol and dry ice. Similar results were reported by Hoffman and Vecchio for AA6061 by welding under water [9]. Temperature history plays a significant role in determining properties within a friction stir weld; therefore, accurate measurement of temperature inside the stir zone during the welding process is crucial if the process is to be understood. Unfortunately, temperature measurement inside the FSW process zone is highly problematic. Several factors make it nearly impossible to capture 0921-5093/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2009.10.039 Please cite this article in press as: P. Upadhyay, A.P. Reynolds, Mater. Sci. Eng. A (2009), doi:10.1016/j.msea.2009.10.039 G Model MSA-25619; 2 No. of Pages 7 ARTICLE IN PRESS P. Upadhyay, A.P. Reynolds / Materials Science and Engineering A xxx (2009) xxx–xxx the actual temperature transient experienced by the weld nugget material. Steep temperature gradients, finite thermocouple size, deformation, and movement of the material in which the thermocouple is embedded combine to make accurate temperature determination complicated [10]. Due to these reasons the peak temperature inside the stirred zone has been estimated using indirect analytical methods like comparison of micro-hardness, nugget grain size, and quantification of precipitates in the matrix [11,12]. Some investigators have used computational models, normally validated with far-field temperature measurements, to deduce the temperature history [11–15]. Numerous studies have been devoted to understand the relationships between properties and welding parameters for 7XXX series alloys [1,2,12,15–18]. These alloys are precipitation hardened with ␩′ (Mg2 Zn) as the primary strengthening precipitate phase. When solution heat treated and subsequently aged using appropriate times and temperatures, the alloy microstructure will have a distribution of fine particles in the solid solution matrix with optimum tensile and other mechanical properties. During the welding process this optimized microstructure is altered at various levels because of the temperature cycle. This cycle will always produce some level of dissolution of precipitates, diffusion of solute, and increase in vacancy concentration. Thus some non-strengthening phases may form and some strengthening precipitates may be coarsened or dissolved, causing varying degrees of loss of strength. The extent of this modification in microstructure will be primarily governed by the temperature history which in turn is dependent on welding parameters used and the thermal boundary conditions during the weld. To obtain the best possible properties in the weld nugget zone it is desirable that the nugget reach a peak temperature that is above the solution heat treatment temperature for the alloy being welded. The nugget which reaches such a peak temperature would be in a condition similar to that obtained after solution heat treatment and quenching; therefore, reprecipitation of strengthening phases may occur during post-weld cooling and subsequent natural and artificial aging treatments. Typically, if a weld is made in a precipitation hardening aluminum alloy and the peak temperature is greater than the solution heat treatment temperature for the alloy, a characteristic “W” shaped hardness distribution is observed. This arises due to solution heat treatment of the nugget and overaging of the heat affected zone (HAZ) [15–17]. If the weld is performed at relatively low power (with a stir zone peak temperature less than approximately 350 ◦ C) the characteristic W shape in the hardness profile will not be observed. With a peak weld temperature near 350 ◦ C, normally, the hardness in the HAZ and the nugget will be similar to each other and less than a T6 or T7 base metal. Since the kinetics of precipitate coarsening are maximum near 350 ◦ C for 7050 and other 7XXX series alloys, the rate of formation of nonstrengthening ␩ phase is at peak near this temperature [19]. During its formation, ␩ phase particles take away a significant amount of solute from the matrix which otherwise would have been available for reprecipitation of strengthening ␩′ phase during the post-weld heat treatment. The situation will be aggravated when the welding speed is lower. Lower welding speed will cause the stir zone to remain in the ␩ phase formation temperature range for a relatively longer time causing the solute depletion to increase. In general, the depth of the HAZ and/or nugget hardness minimum will depend strongly on the time spent near 350 ◦ C without subsequent solution heat treatment, hence, the depth of the minimum is highly dependent on the welding speed which primarily governs the temporal length of the temperature transient [2,20]. In this work we present thermal behavior, torque requirements, and resulting mechanical properties of friction stir welds in alloy 7050 performed in-air, under-water, and under sub-ambient temperature conditions for a wide spectrum of welding speed and rotation rate. The effects of welding parameters and some thermal boundary conditions on the resulting weld properties, in particular, the grain size, hardness distributions, and transverse tensile strength are reported and discussed. 2. Materials and experimental procedure All welding was performed on 6.35 mm thick plates of the high strength aluminum alloy, 7050-T7451 having a nominal composition of 5.6%Zn–2.5%Mg, 1.6%Cu, 0.23%Cr, balance Al. The typical ultimate strength of the alloy in the T7451 temper is 524 MPa. The incipient melting temperature for homogenized 7050 is 488 ◦ C and the solution treatment temperature is 477 ◦ C. The dimensions of the welded plates were 6.35 mm thick, and 101.6 mm wide (yielding a total welded width of 203.2 mm). Welds were produced on an MTS FSW Process Development System (PDS) using Z-axis force control. The welding direction was parallel to the plate rolling direction and the tool rotation axis was normal to the plane of the plate. Welds were made with the parent metal plates in lab air or completely submerged in water: for underwater welds, the depth of water was approximately 25 mm above the top surface of the plates (see Fig. 1). For three sets of parameters, welds were also performed with the parent metal plates submerged in a mixture of 50% ethylene glycol, 50% water and 30 lb of dry ice resulting in a steady far field plate temperature of approximately −25 ◦ C during the weld. . For the sake of convenience, these three conditions will henceforward be referred to as IA (in-air), UW (under-water), and SA (sub-ambient). The combinations of tool rotation speed, welding speed, and Z-axis force used are tabulated in Table 1. The double entries in columns with 300 and 400 rpm for IA and UW welds are because of the use of two welding speeds for each of those rotational speeds. The tool used for production of all welds was of a two piece design with a 17.8 mm diameter single scroll H13 tool steel shoulder and a probe fabricated out of MP-159 (a high temperature cobalt based super alloy) in the shape of a truncated cone (8◦ taper) with threads and three flats. The probe was 6.1 mm long, with a diameter of 7.9 mm at the intersection with the shoulder. Temperature during welding was monitored and recorded using a thermocouple spot welded into the probe on the axis of rotation at the probe mid-plane height. For some select welds, thermocouples were also placed at approximate location of the HAZ hardness minimum on the advancing side of the weld at the mid-plane height. This corresponds to a distance of 5.5–7 mm away from the weld centerline. Samples for metallographic and hardness measurement were ground, polished and etched using Keller’s reagent. Micrographs were obtained from the nugget center. Grain size was measured at the nugget center using the MLI method. Three views at a magnification of 500× were examined. Intersections were counted on a test line of length 100 mm. Five test lines per view were used. Fig. 1. Under-water friction stir welding. Please cite this article in press as: P. Upadhyay, A.P. Reynolds, Mater. Sci. Eng. A (2009), doi:10.1016/j.msea.2009.10.039 ARTICLE IN PRESS G Model MSA-25619; No. of Pages 7 P. Upadhyay, A.P. Reynolds / Materials Science and Engineering A xxx (2009) xxx–xxx 3 Table 1 Welding control variables. Rotational speed (rpm) Welding speed (mm/s) In-air Z force (N) Under-water Z force (N) Sub-ambient Z force (N) 150 1.7 25,800 35,586 200 2.54 25,800 35,586 250 3.4 26,690 34,696 300 3.4, 6.8 24,021, 33,807 31,138, 39,145 35,586, – Hardness and tensile tests were performed after post-weld heat treatment of weld samples at 121 ◦ C for 24 h preceded by a week of natural aging: this approximates an industrial T6 heat treatment (peak strength). A Vickers hardness indenter, with a load of 1000 g and a load application time of 10 s was used for measurement of hardness as a function of distance from the weld centerline on transverse cross-sections along the plate mid-plane. Full thickness (6.35 mm) rectangular transverse tensile samples of 180 mm gauge length and 12.5 mm width were machined from completed welds. Tensile tests were performed using an initial strain rate of 0.0001 s−1 . Three samples were tested for each welding condition. 3. Results and discussion 3.1. Torque and probe temperature Fig. 2 shows the representative thermal history from the thermocouple welded into the probe mid-plane during a two parameter set weld: 400 rpm, 5.1 mm/s and a 540 rpm, 6.8 mm/s. The probe temperature is observed to reach a reasonably steady state condition and to react quickly to changes in the welding parameters (the weld parameters were changed at approximately 140 s of welding time). The temperature measured inside the probe is some average of the temperatures of the material in contact with the probe and does not represent the maximum temperature; however, this measured temperature is quite repeatable [21] and provides an excellent representation of the trends in nugget temperature with changing weld parameters. The maximum measured probe temperature and average measured torque have been graphed against the rpm in Fig. 3 for both IA and UW welds. Note that welding speed is not a constant for this graph (see Table 1). With increasing rotational speed, probe temperature increases and the torque decreases. Two general observations can be made for both the thermal conditions. (1) Peak probe temperature is inversely correlated with the measured Fig. 2. Representative temperature transients from the thermocouple welded into the probe mid-plane during a two parameter set weld: 400 rpm, 5.1 mm/s and 540 rpm, 6.8 mm/s. Parameter switch takes place approximately at 140 s. 400 3.4, 5.1 22,241, 25,800 29,359, 34,252 –, 40,479 540 6.8 28,024 39,145 41,369 650 6.8 30,693 39,145 800 6.8 33,362 40,479 1,000 10.2 37,810 43,593 torque. (2) Both the quantities approach a limiting value as rpm is increased. This may be rationalized in the way proposed originally by Tang et al. [22]: the higher the rotational speed the higher will be the temperature causing the material flow stress to decrease. This decrease in flow stress in turn will limit the power generation by plastic dissipation and, hence, temperature increases. A similar trend in peak temperature was reported by Sato et al. for alloy 6063 in the rpm range of 800–3600 [16]. Yan et al. [20] and Long et al. [23] have also reported similar observations based on experiments and CFD simulations. Considering Fig. 3 again, all other things being equal, the peak probe temperature is consistently lower for UW welds. This is due to enhanced heat transfer from the tool and plate into the surrounding water. The torque and hence power requirement for otherwise equivalent UW welds is higher than that for IA welds. That the power is higher while the peak temperature is lower is a reflection of the relatively higher heat transfer to the environment for the UW welds as compared to the IA welds. The higher torque observed for under-water welding is no doubt related to lower temperature of the material in contact with the tool. The lower temperature will correspond to higher flow stress, thus the tool would require more torque (and, for a given rpm, more power) to “stir” the material. 3.2. Thermal cycle Representative weld thermal cycles for IA, UW, and SA weld conditions for two different welding speeds are shown in Figs. 4 and 5. The data in these figures are obtained from thermocouples placed on the advancing side of the weld at the respective locations of minimum hardness which were determined from hardness testing of identical welds made previously. There are several features of note: firstly, the peak temperature measured in the HAZ minimum hardness location is close to 350 ◦ C for all six conditions. This is consistent with the temperature associated with maximum overaging kinetics for 7XXX type alloys [19,24]. In Fig. 4 (weld made at 3.4 mm/s), it can be observed that the heating rate for the IA and UW welds are similar, but the UW weld experiences a faster cooling rate than the IA weld. The temperature transient for the SA weld Fig. 3. Measured torque and peak probe temperature vs. rpm for under-water and in-air welds for several welding sets. Please cite this article in press as: P. Upadhyay, A.P. Reynolds, Mater. Sci. Eng. A (2009), doi:10.1016/j.msea.2009.10.039 G Model MSA-25619; 4 No. of Pages 7 ARTICLE IN PRESS P. Upadhyay, A.P. Reynolds / Materials Science and Engineering A xxx (2009) xxx–xxx Fig. 4. Thermal history of 3.4 mm/s welds for in-air, under-water and sub-ambient conditions. is different from both the IA and UW welds. Firstly, the initial and final temperatures are significantly lower for the SA weld and secondly, the temporal length of the temperature transient is shorter. The same types of data are shown in Fig. 5 except that the welding speed is 6.8 mm/s. Here, the UW and SA welds exhibit very similar temperature transients except at the low temperatures which have essentially no effect on properties: the initial temperature for the SA weld transient is approximately −25 ◦ C. The temporal lengths of the UW and SA weld temperature transients are much shorter than that of the IA weld. The heating portions of the transients for all three welds are quite similar and the cooling rates for the UW and SA welds are much higher. The temporal length of the temperature transients are compactly represented in Fig. 6 which is a plot of the length of time spent above 200 ◦ C for each of nine welding conditions: UW, IA, and SA at three welding speeds (3.4 mm/s, 5.1 mm/s, and 6.8 mm/s). The choice of 200 ◦ C is arbitrary. From the graph it is seen that, for the three welding conditions shown, the time above 200 ◦ C at the HAZ minimum hardness location is much greater for the IA welds while the differences between the UW and SA welds are small. Also there is hardly any decrease in the time of stay beyond 5.1 mm/s of welding speed. This indicates that there exists a finite limit to the highest cooling rate achievable for any given welding condition. This idea can be illustrated by considering a highly simplified Fig. 5. Thermal history of 6.8 mm/s welds for in-air, under-water and sub-ambient conditions. Fig. 6. Time spent above 200 ◦ C at HAZ vs. welding speed for in-air, under-water and sub-ambient conditions. model of welding in which a point heat source moves across the surface of a thick plate at a given speed. The temperature field solution for this model was first presented by Rosenthal [25] according to which the temperature at any given point on the plate is given by T = T0 + 1 Q · · exp 2k r  −V (r − x)  2˛ where T0 is initial plate temperature, Q is total heat produced by the source, r is the radial distance of the point from the source, V is the welding velocity, ˛ is the thermal diffusivity of the plate and x is the coordinate in the direction of welding. Using an appropriate power value, Q, from a typical FSW and the thermal diffusivity of AA7050 at room temperature, temperature fields were obtained for a series of welding speeds. The times above 200 ◦ C are plotted with respect to welding speeds in Fig. 7. The time spent above 200 ◦ C is observed to follow a power law relationship with the welding speed thus illustrating the diminishing returns in the increase in cooling rate with the increase in the welding speed. 3.3. Hardness Representative hardness distributions after post-weld aging for the three thermal boundary conditions are shown in Fig. 8a. The Fig. 7. Time spent above 200 ◦ C vs. welding speed obtained using the Rosenthal’s solution for a moving heat source. Please cite this article in press as: P. Upadhyay, A.P. Reynolds, Mater. Sci. Eng. A (2009), doi:10.1016/j.msea.2009.10.039 G Model MSA-25619; No. of Pages 7 ARTICLE IN PRESS P. Upadhyay, A.P. Reynolds / Materials Science and Engineering A xxx (2009) xxx–xxx 5 Fig. 8. (a) PWHT micro-hardness trends for in-air, under-water, sub-ambient welds made at 800 rpm, 6.8 mm/s and (b) PWHT micro-hardness distribution for 200 rpm, 2.54 mm/s weld for in-air and under-water. hardness distributions shown are for welds made at 6.8 mm/s and 540 rpm. All of the distributions exhibit the “W” shape typical of 7XXX type alloys welded under conditions for which the nugget temperature is close to the solution heat treatment temperature. The nugget regions exhibit hardness which is equivalent to a T6 temper and the base metal regions exhibit T7 (slightly overaged) hardness. In the HAZ, substantial overaging leads to a hardness minimum. The shape indicates that the nugget region is in a nearly solution heat treated condition prior to post-weld aging for this set of welding parameters. The hardness minima in the SA and UW welds are not as deep as that observed in the IA weld, consistent with the differences in the temporal length of the temperature transient as shown in Fig. 6. Also, the near equivalence between the minimum hardness levels in the UW and SA welds is consistent with the similar lengths of time spent above 200 ◦ C at the HAZ minimum locations. Fig. 8b illustrates an interesting effect of the submerged welding for relatively lower power welds. Here, it can be seen that, for IA and UW welds made at 200 rpm and 2.54 mm/s very different hardness distributions are observed. The UW weld shown in Fig. 8b exhibits a flat hardness profile through the nugget while the IA weld exhibits a W shaped distribution. Compared to the profiles shown in Fig. 8a, produced at higher rpm and weld power, the peak hardness in the IA weld made at 200 rpm is less. Referring to Fig. 3, it is seen that the probe temperature for the 200 rpm IA weld is near 400 ◦ C while the probe temperature for the 200 rpm UW weld is near 350 ◦ C. Hence, in the UW weld, overaging takes place throughout the nugget and in the IA weld, some dissolution followed by reprecipitation occurs in the nugget. The solution treatment is not complete in the IA nugget; therefore the hardness is less than in the welds shown in Fig. 8a. On the other hand, the time for overaging of the UW nugget (Fig. 8b) is less than that for the IA HAZ leading to a higher minimum hardness in the UW weld as compared to the IA weld. Fig. 9 shows the trends of average nugget hardness with measured probe temperature for IA, UW, and SA welds. The average nugget hardness is taken at the plate mid-plane over the distance from edge to edge of the recrystallized region in the weld. Nugget hardness generally increases with the increase in peak probe temperature; however, the increase is not linear. At the lowest probe temperatures the nuggets are apparently in an overaged condition and the hardness is similar to that observed in the HAZ minima (near HVN = 120). In the intermediate temperature range, the nugget hardness rises rapidly with increasing temperature, presumably because post-weld, there is a greater amount of solute available for reprecipitation. In the highest temperature range, the effect of increasing temperature on hardness diminishes because above some temperature, the additional solute obtained by further temperature increase becomes insignificant. For a given probe temperature, the UW and SA welds generally exhibit higher nugget hardness after PWHT. This can be rationalized on the basis of a higher quench rate in the UW and SA welds which will serve to trap more solute and more vacancies in the as welded nugget, thereby enhancing the response to the post-weld aging treatment. Fig. 10 shows the trends in HAZ minimum hardness for the SA, UW and IA welds. The minimum hardness values are plotted vs. the welding speed. It has previously been shown that the HAZ minimum hardness in precipitation hardening aluminum alloys is almost solely dependent on the welding speed [2,20]. The UW and SA minimum hardness is consistently higher than that of IA welds made with the same welding speed: as for the nugget hardness values, this is most likely due to increased quench rates in the UW and SA welds (see Fig. 6). The increase in minimum hardness with increasing welding speed is not linear. The increase is rapid at low welding speeds and then slows substantially as welding speed is further increased. For the UW welds, almost no increase in minimum hardness is observed above a welding speed of 5.1 mm/s. This observation is consistent with the data in Fig. 6 which indicate that the time spent above 200 ◦ C in the minimum hardness location is not reduced substantially by welding at ever increasing rates and Fig. 9. Average nugget hardness vs. peak probe temperature. Please cite this article in press as: P. Upadhyay, A.P. Reynolds, Mater. Sci. Eng. A (2009), doi:10.1016/j.msea.2009.10.039 G Model MSA-25619; No. of Pages 7 6 ARTICLE IN PRESS P. Upadhyay, A.P. Reynolds / Materials Science and Engineering A xxx (2009) xxx–xxx Fig. 10. HAZ minimum hardness vs. welding speed. with the results of the Rosenthal analysis presented in Fig. 7. This is a useful observation from a practical standpoint, indicating that little improvement in weld properties is likely to arise from great effort to increase the welding speed above some level. Fig. 12. Ultimate tensile strength against tool rpm for in-air and under-water cases. 3.5. Tensile strength and correlation with weld hardness distributions The nugget grain size is plotted against peak probe temperature with corresponding welding speed indicated by symbols in Fig. 11. Notice that at the welding speed of 6.8 mm/s welds were performed at three rotational speeds, hence three data points. The arrows indicate corresponding grain sizes for in-air and underwater welds made with the same parameters. The grain sizes of welds performed below 400 rpm, with probe temperatures below 400 ◦ C, were not resolvable by optical means and are not reported here. For the rest of the weld parameters the UW grain size is consistently smaller than the IA. This can be attributed to the lower peak temperature in the UW welds. This should be true whether one takes the stance that grain size is due to post-weld recrystallization and growth [1] or due to geometric dynamic recrystallization [26–28] or continuous dynamic recrystallization [18,29]. Fig. 12 is a plot of the ultimate tensile strength (UTS) in transverse tension vs. tool rotation rate for the IA and UW cases. Error bars indicate the total range of test results for three specimens per condition. Generally, the UW welds are substantially stronger than the corresponding IA welds. At 200 rpm (welding speed of 2.54 mm/s), the UW weld is weaker than the corresponding IA weld. This is the only situation in which any of the welds without a defect broke in the nugget: the hardness distribution for the 200 rpm UW weld (shown in Fig. 8b) is essentially flat in the HAZ and nugget areas leading to failure through the nugget. It is evident from examination of all the transverse tensile data that the rpm cannot fully correlate the transverse tensile strength. Fig. 13 shows average tensile strength plotted against the measured HAZ minimum hardness on the retreating side of the weld for those welds which exhibited a “W” shaped hardness distribution. Obvious and well known direct correlation between tensile strength and the minimum hardness is exhibited by all IA, UW and SA welds. Fig. 14 shows % elongation in transverse tension (solid symbols) and the difference between the average nugget hardness and the HAZ minimum values (VHN, open symbols) both plotted against the probe temperature. The solid and dotted lines are fit by eye Fig. 11. Average nugget grain size vs. peak probe temperature for different welding speeds. The arrows show equivalent welds done in air and under water. Fig. 13. Correlation between ultimate tensile strength and HAZ minimum hardness in the retreating side of the weld. 3.4. Nugget grain size Please cite this article in press as: P. Upadhyay, A.P. Reynolds, Mater. Sci. Eng. A (2009), doi:10.1016/j.msea.2009.10.039 G Model MSA-25619; No. of Pages 7 ARTICLE IN PRESS P. Upadhyay, A.P. Reynolds / Materials Science and Engineering A xxx (2009) xxx–xxx 7 (2) While not measured, increased cooling rates in the weld nugget are implied by the higher hardness observed for UW vs. IA welds with equivalent probe temperature (Fig. 9). (3) Increased cooling rates in the HAZ due to enhanced convection result in higher HAZ minimum hardness values for UW and SA welds vs. otherwise equivalent IA welds. (4) Transverse tensile strength is enhanced by UW welding except for welding conditions for which the UW weld temperature is too low to enable substantial reprecipitation in the weld nugget during PWHT. The transverse tensile strength exhibits a strong, direct, correlation with the HAZ minimum hardness. (5) Temperature measurement in the HAZ minimum hardness region indicates that, for the experimental setups under consideration, little reduction in time at temperature can be achieved by welding at speeds greater than 5.1 mm/s. (6) Under the conditions studied, pre-cooling to −25 ◦ C does not provide a significant benefit compared to welding under water at ambient temperature. Acknowledgements Fig. 14. Percentage elongation and difference of hardness values between nugget and HAZ minimum plotted against the peak probe temperature. The dotted and solid lines are fit by eye. to the elongation and VHN data respectively. Firstly, it can be noted that the elongation values are similar for the IA and UW conditions. Elongation is relatively high at low values of probe T and declines to a plateau value near 6% at probe T > 400 ◦ C. The VHN increases with increasing probe T, from a value near zero for a probe T of 320 ◦ C to a high of near 60. For probe T substantially greater than 350 ◦ C, precipitate dissolution in the nugget is occurring simultaneously with coarsening in the HAZ leading to higher nugget hardness and lower HAZ hardness after PWHT. One side effect of high nugget hardness is that deformation during transverse tensile testing tends to be concentrated in the HAZ. As the VHN becomes larger and larger, the strain concentration in the HAZ becomes more prominent and the % elongation declines. Once the VHN reaches a high enough value, essentially no deformation occurs in the nugget during transverse loading: here the critical value of VHN is between 35 and 40. The % elongation is not actually a reflection of the ductility of the various weld regions: it is related to variations in strength which cause strain localization. 3.6. Summary and conclusions In this study the effects of changing the thermal boundary condition and the FSW control variables on weld properties and some FSW response variables have been examined. Several general trends have been observed and some useful correlations provided. Specifically: (1) All other things being equal the increased surface convection resulting from welding under water compared to in-air welding results in: (a) Reduced probe temperature. (b) Increased torque and power consumption. (c) Decreased nugget grain size. (d) Increased cooling rates in the HAZ. The authors acknowledge the financial support of the Center for Friction Stir Processing which is a National Science Foundation I/UCRC supported by Grant No. EEC-0437341. The authors thank Dr. Wei Tang and Daniel Wilhelm, Department of Mechanical Engineering, University of South Carolina, Columbia, SC, USA for their help in preparing the weld joints. References [1] Y.S. Sato, H. Kokawa, M. Enomoto, S. Jogan, T. Hashimoto, Metall. Mater. Trans. A 30A (December) (1999) 3125–3130. [2] A.P. Reynolds, W. Tang, Z. Khandkar, J.A. Khan, K. Linder, Sci. Technol. Weld. Join. 10 (2) (2000) 190–199. [3] T.W. Nelson, R.J. Steel, W.J. Arbegast, Sci. Technol. Weld. Join. 8 (4) (2003) 283–288. [4] M. Rowe, S. Liu, Sci. Technol. Weld. Join. 6 (6) (2001) 387–396. [5] S. Benavides, Y. Li, L.E. Murr, D. Brown, J.C. McClure, Scripta Mater. 41 (8) (1999) 809–881. [6] D. Sakurada, K. Kathoh, H. Tokisue, J. Jpn. Inst. Light Met. 52 (2) (2002) 2–6. [7] P. Staron, M. Kocak, S. Williams, Appl. Phys. A 74 (Suppl.) (2002) S1161–S1162. [8] J.-Q. Su, T.W. Nelson, C.J. Sterling, J. Mater. Res. 18 (August (8)) (2003). [9] D.C. Hoffman, K.S. Vecchio, Mater. Sci. Eng. A 402 (2005) 234–241. [10] R. Nandan, T. DebRoy, H.K.D.H. Bhadeshia, Prog. Mater. Sci. 53 (2008) 980–1023. [11] M.W. Mahoney, C.G. Rhodes, J.G. Flintoff, R.A. Spurling, W.H. Bingel, Metall. Mater. Trans. A 29A (1998) 1955–1964. [12] G. Liu, L.E. Murr, C.S. Niou, J.C. McClure, F.R. Vega, Scripta Mater. 37 (1997) 355. [13] H. Schmidt, J. Hattel, Sci. Technol. Weld. Join. 10 (2) (2005) 176–186. [14] P.A. Colegrove, H.R. Shercliff, R. Zettler, Sci. Technol. Weld. Join. 10 (2) (2007) 176–186. [15] K.V. Jata, K.K. Sankaran, J.J. Ruschau, Metall. Met. Trans. A 31A (2000) 2181–2192. [16] Y.S. Sato, M. Urata, H. Kokawa, Metall. Mater. Trans. A 33 (3) (2002) 625–635. [17] K.A.A. Hassan, P.B. Prangnell, A.F. Norman, D.A. Price, S.W. Williams, Sci. Technol. Weld. Join. 8 (4) (2003) 257–268. [18] J.-Q. Su, T.W. Nelson, R. Mishra, M. Mahoney, Acta Mater. 51 (2003) 713–729. [19] O. Frigaard, O. Grong, O.T. Midling, Metall. Met. Trans. A 32A (2001) 1189–1200. [20] J. Yan, M.A. Sutton, A.P. Reynolds, Sci. Technol. Weld. Join. 10 (6) (2005) 725–736. [21] R. Brown, W. Tang, A.P. Reynolds, Mater. Sci. Eng. A513–514 (2009) 115–121. [22] W. Tang, X. Guo, J.C. McClure, L.E. Murr, J. Mater. Process. Manuf. Sci. 7 (1998) 163–172. [23] T. Long, W. Tang, A.P. Reynolds, Sci. Technol. Weld. Join. 12 (4) (2007) 311–317. [24] S.D. Liu, X.M. Zhang, J.H. You, Z.B. Huang, Z.P. Zhou, Mater. Sci. Technol. 24 (12) (2008) 1419–1421. [25] D. Rosenthal, Weld. J. 20 (1941) 220. [26] K.V. Jata, S.L. Semiatin, Scripta Mater. 43 (2000) 743. [27] Y. Li, E.A. Trillo, L.E. Murr, J. Mater. Sci. Lett. 19 (2000) 1047. [28] Z.Y. Ma, R.S. Mishra, M.W. Mahoney, Acta Mater. 50 (2002) 4419. [29] B. Heinz, B. Skrotzki, Metall. Mater. Trans. B 33 (6) (2002) 489. Please cite this article in press as: P. Upadhyay, A.P. Reynolds, Mater. Sci. Eng. A (2009), doi:10.1016/j.msea.2009.10.039