Journal of Materials Science & Technology  2019 , 35 (7): 1412-1421 https://doi.org/10.1016/j.jmst.2019.01.018

Orginal Article

Evaluation of dynamic development of grain structure during friction stir welding of pure copper using a quasi in situ method

X.C. Liuab*, Y.F. Sunac, T. Nagiraa, K. Ushiodaa, H. Fujiia**

aJoining and Welding Research Institute, Osaka University, 11-1 Mihogaoka, Ibaraki, Osaka, 567-0047, Japan
bSchool of Materials Science and Engineering, Northwestern Polytechnical University, Xi’an, 710072, China;
cSchool of Materials Science and Engineering, Zhengzhou University, Zhengzhou, 450001, China

Received: 2018-11-29

Revised:  2019-01-19

Accepted:  2019-01-20

Online:  2019-07-20

Copyright:  2019 Editorial board of Journal of Materials Science & Technology Copyright reserved, Editorial board of Journal of Materials Science & Technology

More

Abstract

By employing a quasi in situ method, we investigated the dynamic evolution of the grain structure considering the material flow, strain, and strain rate in the friction stir welding of pure copper. The tool ‘stop action’ and rapid cooling were employed and a brass foil was used as a marker to show the material flow path. The grain structure along the material flow path was characterised using electron backscatter diffraction. Static recrystallization occurs for the work-hardened base material in the preheating stage in front of the tool. In the acceleration flow stage, grains are significantly refined by plastic deformation, discontinuous dynamic recrystallization, annealing twinning during the strain-induced boundary migration and slight continuous dynamic recrystallization. In the deceleration flow stage, due to a strain reversal, the grain first coarsens, and is thereafter refined again. Finally, the hot-deformed material in the shoulder-affected zone is ‘frozen’ directly whereas that in the probe-affected zone undergoes significant annealing; thus, the recrystallized microstructure and 45°-rotated cube texture are obtained in the probe-affected zone.

Keywords: Friction stir welding ; Grain structure ; Material flow ; Pure copper ; EBSD

0

PDF (12456KB) Metadata Metrics Related articles

Cite this article Export EndNote Ris Bibtex

X.C. Liu, Y.F. Sun, T. Nagira, K. Ushioda, H. Fujii. Evaluation of dynamic development of grain structure during friction stir welding of pure copper using a quasi in situ method[J]. Journal of Materials Science & Technology, 2019, 35(7): 1412-1421 https://doi.org/10.1016/j.jmst.2019.01.018

1. Introduction

Friction stir welding (FSW) is well known as a solid-state joining technique [1,2] that can produce good joints by avoiding hot cracking, porosity, etc. in contrast to fusion welding [3]. Until recently, due to the complex material flow in FSW, it has been difficult to establish a relationship between the processing parameters (tool rotation speed, welding speed, and tool geometrical shape) and the physical metallurgical parameters (strain, strain rate, and deformation temperature) [4], leading to difficulties in control of the weld properties. Many studies [[5], [6], [7]] treated FSW as a ‘black box’ to investigate the relationship between the processing parameters and the final microstructure and mechanical properties, and established an empirical model for production. Some studies [[8], [9], [10], [11], [12], [13], [14]] investigated the microstructural evolution during FSW. However, limited by the knowledge of the material flow during FSW, they investigated the microstructural evolution on the weld transverse cross-sections [[8], [9], [10]], around the keyhole [[11], [12], [13]], or along the weld length direction [9,10,14]. However, these investigations were not along the true material flow path. Furthermore, they were only phenomenological and did not establish quantitative relationships between the physical metallurgical parameters and the microstructural evolution. For examples, Mironov et al. [10] concluded that the microstructural evolution during FSW of pure copper was dominated by discontinuous dynamic recrystallization at temperatures above $\widetilde{0}$.5 Tm but continuous dynamic recrystallization at temperatures below $\widetilde{0}$.5 Tm (Tm is the melting point) during the deformation stage, and the microstructure coarsening as well as the material softening primarily occurred during weld cooling cycle. Xu et al. [12] reported that the microstructure in the stir zone was formed by grain subdivision during the deformation stage and emphasized that continuous and discontinuous static recrystallizations occurred during the cooling stage. However, the variation in the grain size with respect to the strain and strain rate during the deformation stage is not yet clear. The difference in the microstructure evolution between the shoulder-affected zone and the probe-affected zone is also unclear.

Some studies on the material flow of aluminium alloy FSW showed that a rotation flow zone may surround the upper part of the probe [[15], [16], [17]]. The material in the rotation flow zone travels with the tool for a long distance and is finally deposited near the top surface of the weld [17]· A similar phenomenon has also been observed during the FSW of copper alloys. As shown in Fig. 1, Cu90Zn10 and copper plates were arrayed side-by-side, and then the FSW tool welded from the Cu90Zn10 to the copper. It can be seen that a lot of Cu90Zn10 surrounded the probe, forming an inverted cone rotation flow zone. The copper in front of the tool bypassed the rotation flow zone and was directly transferred to the rear side of the tool, forming most of the final weld. Only a little of the weld near the weld top surface was affected by the rotation flow zone. Therefore, it is believed that the rotation flow zone does not involve the main weld microstructure evolution. However, these investigations [[8], [9], [10], [11], [12], [13], [14]] did not exclude this zone. Therefore, their conclusions should be accepted with reservations.

Fig. 1.   Horizontal and longitudinal cross-sections of the keyhole showing the rotation flow zone by the tool traversing two dissimilar copper alloys (800 rpm-100 mm/min).

The microstructural evolution depends on not only the true material flow path, but also on the physical metallurgical parameters (strain, strain rate, and temperature). During the viscoplastic material flow in FSW, the accumulative strain must increase continuously, but the strain rate may increase or decrease, depending on the material flow state [4,18,19]. The deformation temperature is also an important factor for the microstructural evolution [9,10]. In spite of the severe temperature gradient at the border of the stir zone (SZ), under steady-state welding conditions, the temperature within the SZ is relatively uniform [15,[20], [21], [22]]. This is easy to understand because the continuous convective heat transfer is caused by the material rotation flow in the SZ. However, unlike in other plastic deformation methods, the strain, strain rate, and deformation temperatures during FSW are usually difficult to obtain directly. The authors recently developed experimental methods to evaluate the strain and strain rate during the FSW [23], which make it possible to investigate the effect of the strain and strain rate on the grain structure development during FSW.

In this study, by employing the method described in [23], the material flow, strain, and strain rate were combined with the microstructural evolution. Pure copper, which has a face-centred cubic (fcc) structure and relatively low stacking fault energy (SFE), was used as the material for the experiments. Electron backscatter diffraction (EBSD) investigation of the microstructure along the true material flow path was conducted on the horizontal cross-section of the keyhole. In order to trace the material flow path, the marker insert method [18,24] and metallographic analysis were adopted jointly. A ‘stop action’ of the FSW tool and an in-process rapid cooling method were adopted to ‘freeze’ the transient microstructure of the SZ, providing a static snapshot of the dynamic microstructural change for the experimental observations. The strain and strain rate were calculated according to the distortion of the marker material. The dynamic development of the grain structure evaluated using this quasi in situ method are discussed in detail considering the strain, strain rate, and the effect of the preheating and cooling stages along specific flow lines through the weld.

2. Experimental

In this experiment, 3-mm-thick commercial pure copper C1100P in the 1/4H condition was used as the base material. Fig. 2 shows a schematic of the experiment. During the rapid cooling FSW, liquid CO2 was rapidly jetted onto the fresh weld surface through two tandem pipes and was moved with the FSW tool simultaneously [25]. Liquid CO2 has a good cooling capacity due to its high specific heat, ultralow temperature and high flow rate. Other cooling media such as water [26,27] also can obtained a very good cooling capacity. Nevertheless, the liquid CO2 can evaporate rapidly once it jets on the weld and therefore the liquid CO2 only suppresses the post-annealing but not decrease the deformation temperature significantly (not like water cooling [26]). A thin foil of 60/40 brass (0.5 mm nominal thickness) was inserted into the butting surface of the two workpieces as a marker. 60/40 brass has a lower flow stress than pure copper at 650 °C (approximately equal to welding peak temperature) and a strain rate of 1 s-1 [[28], [29], [30]]. This indicates that the 60/40 brass can deform compatibly together with the pure copper under the FSW conditions. The tool ‘stop action’ was applied using the emergency stop, which indicates that the FSW machine lost power suddenly and the tool was pulled out immediately. Note that the liquid CO2 jetting was still active for a few seconds to ensure that the keyhole was fully cooled.

Fig. 2.   Schematic of the experiment.

A tungsten-carbide-based alloy tool containing a concave shoulder of 12 mm diameter and a smooth cylindrical probe of 2.8 mm length and 4 mm diameter was used. The tilt angle of the tool was 3°. The FSW machine was operated in a displacement control mode. The welding and rotating speeds were 150 mm/min and 800 rpm, respectively, which are typical parameters for the FSW of copper using liquid CO2 cooling [12]. The thermal cycles were measured at the surface of the workpiece at a distance of 1 mm from the weld border and the weld bottom.

After welding, the specimen containing the keyhole was removed via wire-cutting. In order to show the potential difference in the microstructural evolution at the upper and lower parts of the weld, the specimen was ground 0.5 mm from the top and bottom using sandpapers in flowing water, defined as the 0.5-mm-plane (upper part) and 2.5-mm-plane (lower part, 0.5 mm from the bottom), respectively. Their metallographic specimens were prepared via mechanical polishing and subsequent etching with a solution of FeCl3 (5 g) + HCl (25 mL) + H2O (75 mL). The metallographic photographs for each plane were obtained using an optical microscope and were spliced to obtain a panoramic view.

For the EBSD measurement, the specimens were electro-polished in a solution of 87.5% orthophosphoric acid at 0 °C and 1.5 V. The measurement was performed every 0.5 mm in the welding direction (WD) and along the flow path of the marker material using a JEM-7001FA instrument with the TSL OIM system at 15 kV. The step size ranged from 0.15 to 0.7 μm, depending on the grain size of the samples. The area used to EBSD analysis contains more than 1000 grains. To avoid spurious boundaries caused by orientation noise, a lower-limit boundary misorientation cut-off of 2° was employed. A criterion of 15° was used to differentiate the low-angle grain boundaries (LABs) and high-angle grain boundaries (HABs). The grain was defined with a grain tolerance of 5° and minimum grain size of 2 pixels. The grain size was quantified by determining the area of each grain in the EBSD map and calculating its circle equivalent diameter. The Σ3 twin boundary (TB) was determined by the <111>/60° axis/angle pair with a tolerance of 5°.

3. Results and discussion

3.1. Material flow, strain and strain rate

Fig. 3 shows the material flow features, strain and strain rate distributions on the keyhole horizontal planes. The strain was evaluated based on the distortion of the marker material. The material flow velocity can be obtained by assuming that the material flow in the passageway formed by the marker material is a one-dimensional incompressible fluid flow. In this way, the strain rate can be calculated according to the strain increment and the deformation time which was calculated by using the average value of the velocities at two adjacent measured points as the average material flow velocity of the corresponding interval. More details about the measurements and calculations of the material flow, strain and strain rate refer to our previous study [23]. In Fig. 3, in order to describe the deformation characteristics at various locations, a uniaxial coordinate system was established. The keyhole centre was set as the origin and the material feed direction was set as the positive direction. The dimensions are in millimetres. On the 0.5-mm-plane (Fig. 3(a)), the material flow is continuous, which is known to occur in the shoulder-affected zone [31,32]. The material undergoes a preheating stage first in front of the deformation zone, then, an acceleration flow stage, a high-velocity flow stage, a deceleration flow stage, and finally an annealing stage behind the deformation zone (Fig. 3(a)). The accumulated strain increases in a two stair-step shape as the material flows from the front side to the rear side of the tool. However, due to the opposite direction of the strain at the acceleration and deceleration flow stages, strain reversal occurs [23]. Therefore, the strain rate varies in a sinusoidal shape, with a positive peak at the acceleration stage and a negative peak at the deceleration stage. On the 2.5-mm-plane, the material first undergoes a preheating stage and an acceleration flow stage, and then flows around the probe surface to the rear side of the probe, forming banded structures (Fig. 3(b)). This feature is usually observed in the probe-affected zone [31,32]. Due to the formation of the banded structures, there is no obvious strain reversal occurring [23]. It can be seen that the average strain rate during the band formation is significantly higher than the maximum strain rate on the 0.5-mm-plane.

Fig. 3.   Material flow features, strain and strain rate distributions on the keyhole horizontal planes [23]: (a) and (c) are the material flow on the 0.5-mm-plane and the corresponding strain, material flow velocity and strain rate, respectively; (b) and (d) are the material flow on the 2.5-mm-plane and the corresponding strain and strain rate, respectively.

3.2. Preheating stage

As shown in Fig. 3, the preheating stage exists both on the 0.5-mm-plane and the 2.5-mm-plane. The measured temperature history indicates that the preheating temperature reached ∼500 °C, and the heating from 100 to 500 °C took ∼10 s. Fig. 4 shows the typical microstructure of the base material and the microstructure in the preheating stage (x = -7, 0.5-mm-plane). The base material consists of coarse grains of 5.23 μm diameter with 60.1% LABs and 11.7% Σ3 TBs. Fig. 4(c) shows that most of the LABs have misorientation angles smaller than 5° and the misorientation angles of the HABs are mainly concentrated at 60°. There is no apparent micro-texture (Fig. 4(d)), due to the annealing and subsequent slight cold-rolling (1/4H condition). After preheating, numerous LAB-free grains can be observed (Fig. 4(f)). Among these LAB-free grains, many annealing twins are formed. In local areas, some grains in which many LABs still exist can be observed. In contrast to the base material, the average grain size slightly increases to 5.48 μm. Moreover, the LAB fraction significantly decreases to 43%, whereas the Σ3 TB fraction increases to 26.1% (Fig. 4(g)). The annealing TBs are usually formed during the migration of the HABs in a low-SFE fcc metal [33]. The (001) pole figure (PF) in Fig. 4h shows that the cube texture component increases due to the preheating. The cube texture is a typical recrystallization texture for fcc metals [34]. These phenomena indicate that some static recrystallization and grain growth occur in the preheating stage for the cold-rolled starting material.

Fig. 4.   Microstructural features of the base material (a-d) and the preheating stage (e-h) (x = -7, 0.5-mm-plane): (a, e) EBSD orientation map; (b, f) grain boundaries map; (c, g) misorientation distribution; (d, h) (001) PF. Note: the welding direction (WD) is the same as the rolling direction (RD) of the base material.

3.3. Acceleration stage

The acceleration stage also exists both on the 0.5-mm-plane and the 2.5-mm-plane. In this stage, the material mainly undergoes plastic deformation and grain refinement at elevated temperature (more than 610 °C). Therefore, the microstructural evolution on the two planes is similar whereas the subjected temperature, strain/strain rate, and the resultant grain size may be different.

The average grain size, LAB fraction, and TB fraction with respect to the accumulative true strain both on the 0.5-mm-plane and 2.5-mm-plane are shown in Fig. 5. The typical grain structures at different strains for the two planes are shown in Figs. 6, and 7, respectively. On the 0.5-mm-plane, the grain size almost progressively decreases with the accumulated true strain (Fig. 5(a)). The LAB and Σ3 TB fractions have opposite variation trends. As shown in Fig. 6a and d, at a small strain (ε=0.06, x=-4), many LABs reoccur in the grain interior, resulting in the increase in the LAB fraction from 49.2% to 67.6% (Fig. 5). Moreover, some TBs begin to lose their ideal twinning relationship, and transform into common grain boundaries, and thus, the Σ3 TB fraction significantly decreases to 6.8%. As the accumulative true strain increases to 0.92 (x = -2, Fig. 6(b) and (e)), the grains are refined to an average size of 2.24 μm, and they have very low or very high LAB densities. Furthermore, the Σ3 TBs mainly occur in the grains with a low LAB density. These phenomena are still observed as the accumulative true strain increases to 1.45 (x = -0.5), as shown in Fig. 6(c) and (f), and the grain size is further refined to 1.72 μm. No deformation twin is observed during this process. These features indicate that plastic deformation via dislocation activities and dynamic recrystallization occurs alternately at the acceleration stage.

Fig. 5.   Variations in the grain size, LABs fraction and TBs fraction with respect to the accumulative true strain at the acceleration stage: (a) 0.5-mm-plane; (b) 2.5-mm-plane. Note: the error bars are based on multiple measurements.

Fig. 6.   Variations in the grain structure on the 0.5-mm-plane at the acceleration stage shown by the EBSD orientation maps (a-c) and grain boundary maps (d-f): (a, d) ε=0.06, x = -4; (b, e) ε=0.92, x = -2; (c, f) ε=1.45, x = -0.5.

Fig. 7.   Variations in the grain structure on the 2.5-mm-plane at the acceleration stage shown by the EBSD orientation maps (a-d) and grain boundary maps (e-h): (a, e) ε=0.05, x = -3; (b, f) ε=0.10, x = -2.5; (c, g) ε=0.32, x = -2; (d, h)ε≈1.03, x =-1.5.

On the 2.5-mm-plane, a similar grain structure evolution is also observed. As shown in Fig. 5(b), the grain size progressively decreases with the increasing strain. The LABs fraction significantly increases at the beginning of the deformation, and then gradually decreases with the strain increase. However, the Σ3 TB fraction significantly decreases at the beginning, and then nearly remains constant. From ε=0.10 (x = -2.5), the recrystallized grains begin to be observed, as shown by the special grains with very few LABs inside in Fig. 7(f)-(h). The minimum grain size reaches 1.32 μm as the accumulative true strain increases above 1.03, which is finer than that on the 0.5-mm-plane although the strain is smaller. This should be attributed to the higher strain rate and the possibly lower temperature at this position [15,[20], [21], [22]]. In addition, the starting grain size on the 2.5-mm-plane (6.05 μm) is slightly larger than that on the 0.5-mm-plane (5.6 μm), indicating that the preheating time is longer on the 2.5-mm-plane. This may be caused by the coverage of the shoulder and the small deformation zone on the 2.5-mm-plane.

Fig. 8 shows several typical mechanisms of the evolution of grain structure at the acceleration stage. As shown in Fig. 8(a)-(c), the HAB bulging can be observed. The kernel average misorientation (KAM) map in Fig. 8(c) shows that there is an apparent difference in the dislocation density on opposite sides of the grain boundary. This is a typical feature of discontinuous dynamic recrystallization (DDRX), which is usually observed during the plastic deformation of low-SFE metals [35]. A similar phenomenon was reported in Refs. [10,12]. In Fig. 8(b), the three marked HABs segments have misorientation angles slightly over 15°. This may be due to the presence of an orientation gradient, or the continuous dynamic recrystallization (CDRX) via accumulation and rearrangement of the dislocations [36,37]. For low-SFE metals, it has been observed that CDRX is dominant during the large plastic deformation at low temperatures [10,38]. DDRX may follow a HAB segment formed by an orientation gradient or CDRX within a grain. As shown in Fig. 8(d)-(f), an HAB segment together with a Σ3 TB exists in the deformed region. The KAM map (Fig. 8(f)) shows that the area surrounded by this HAB segment has a very low dislocation density, and the Σ3 TB is behind the bulged HAB. The low dislocation density areas have misorientation angles of 27° and 60° with respect to the neighbouring deformed grain (Fig. 8(e)). Apparently, the misorientation angle of 60° resulted from the Σ3 TB, whereas the angle of 27° was possibly due to an individual HAB formed by an orientation gradient or the CDRX. During the HAB migration, annealing twins (or Σ3 TBs) can be formed owing to the accidental occurrence of a stacking fault [33]. Here, the migration of the HABs can be regarded as a part of the DDRX [39,40]. In addition, DDRX dominates the recrystallization process, as indicated by the HAB migration frequently observed in Fig. 8(d)-(f). Another evolution mechanism involves the annealing twins. Fig. 8(g)-(i) shows perfect twins evolving into common grains in the grain interior. Due to the continuous plastic deformation, the annealing twins lose their twinning relationship gradually and develop into common grains, as shown by the gradual deviation of the marked misorientation angles from 60°. In this process, sessile Frank partial dislocations and/or sessile unit dislocations formed on the TB induce atomic steps on the TB and lead to the accumulation of gliding dislocations at the TB, which results in the transition from the coherent TB to incoherent grain boundary [41,42]. These new common grain boundaries may become the new origins of the DDRX, thus promoting the grain refinement further.

Fig. 8.   Mechanisms of the grain structure evolution at the acceleration stage: (a, d) IPF maps of the local microstructures; (b, e) the corresponding grain boundary maps and (c, f) KAM maps; (g-i) grain boundary maps showing perfect twins evolving into common grains; in the grain boundary maps, the blue, red and green lines represent HABs, LABs and Σ3 TBs, respectively; in the KAM maps, the black lines represent HABs.

In addition, the DDRX grains usually have different orientations from those of the deformed grains. An example is shown in Fig. 9. The recrystallized grains (Fig. 9(a)) were selected based on the condition of the orientation spread <2° [10,12]. Their (001) PF (Fig. 9(b)) contains two apparent rotated cube orientations. The 001 orientations of these recrystallized grains are arrayed perpendicular to the shear direction (SD) and shear plane normal (SPN), as shown in Fig. 9(c). The deformed grains (Fig. 9(d), orientation spread >2°) present an apparent B/$\bar{B}$ texture [43] (Fig. 9(e)). Almost all the 101 orientations are parallel to the SD (Fig. 9(f)).

Fig. 9.   Orientation relationship between the recrystallized grains and the deformed grains (x = -2.5, 2.5-mm-plane): (a) EBSD image quality map of the recrystallized grains (orientation spread<2°); (b) (001) PF in the SD and SPN; (c) inverse PF in the normal direction; (d) EBSD image quality map of the deformed grains (orientation spread>2°); (e) (111) PF in the SD and SPN; (f) inverse PF in the shear direction; red lines denote LABs, blue lines denote HABs and green lines denote Σ3 TBs.

3.4. High-velocity flow stage

Fig. 10 shows the typical microstructures at the high-velocity flow stage. They are located at x = 0 on the 0.5-mm-plane (Fig. 10(a, c)) and at x = 1 on the 2.5-mm-plane (Fig. 10(b, d)). On the 0.5-mm-plane, the material flow is nearly at a peak velocity, whereas the strain rate is very low, i.e., only 0.04 s-1. The accumulative true strain is 1.45, and the average grain size decreases to 1.62 μm. The grain structure is still characterised by a few recrystallized grains and a large amount of deformed grains containing many disorganised LABs (Fig. 10(a)), resulting in a LAB fraction of 56.8% (Fig. 10(c)). The Σ3 TB fraction is only 6.3% (Fig. 10(c)), which mainly occurs in the recrystallized grains. On the 2.5-mm-plane, as the strain rate is higher and the temperature is possibly lower [15,[20], [21], [22]] than that on the 0.5-mm-plane, the grain is refined to 1.10 μm, but the grain structure is similar to that on the 0.5-mm-plane (Fig. 10(b)). However, due to the larger grain refinement, the HABs are bound to increase, and accordingly, the LAB fraction decreases to 41.8% (Fig. 10(d)). Notably, the grain size on the 2.5-mm-plane from x = -1 to x = 1 is nearly constant. Therefore, it is concluded that the strain is sufficiently high and the strain rate is nearly constant. This agrees well with the experimentally evaluated average strain rate (∼21 s-1) during the banded structure formation [23].

Fig. 10.   Microstructural features at the high-velocity flow stage: (a, c) the 0.5-mm-plane, x = 0; (b, d) the 2.5-mm-plane, x = 1; (a, b) grain boundary map; (c, d) misorientation distribution.

3.5. Deceleration stage

Fig. 11 shows the average grain size, LAB fraction, and Σ3 TB fraction with respect to the accumulative true strain at the deceleration stage of the 0.5-mm-plane. The typical grain structures at different strain are shown in Fig. 12. Significantly different from the acceleration flow stage, the grain size first significantly increases and thereafter slightly decreases with the increase in the accumulative true strain. As shown in Fig. 11, at ε≤1.76, the grain size increases gradually accompanied by the decrease in the LAB fraction and the increase in the Σ3 TB fraction. The microstructure in this interval is shown in Fig. 12(a) and (e). These figures indicate that the recrystallized grains grow, and no significant deformation-induced LAB formation and twin disruption occurs. These results seem to contradict common sense. However, the negative shear deformation in this stage should not be ignored. Due to the deceleration of material flow, the material is gradually accumulated in the material flow path. The shear mode is opposite to that in the acceleration stage. Therefore, the previous approximated tensile strain reverses to an approximated compressive strain [23]. During this process, the grain undergoes significant growth via the strain-induced boundary migration (or DDRX), as shown in Fig. 12(a)-(c) and (e)-(g). Notably, this grain growth is not attributed to the temperature changes because the temperature gradient in the flow zone is negligible [15,[20], [21], [22]]. As the compressed strain is developed to a certain level, the rapidly growing grains are deformed and recrystallized again and thus, the grain size decreases again at ε≥2.54, as shown in Fig. 12(d) and (h).

Fig. 11.   Variations in the statistical grain size, LABs fraction and Σ3 TBs fraction with respect to the accumulative true strain at the deceleration stage on the 0.5-mm-plane. Note: the error bars are based on multiple measurements.

Fig. 12.   Variations in the grain structure at the deceleration stage on the 0.5-mm-plane shown by the EBSD orientation maps (a-d) and grain boundary maps (e-h): (a, e) ε=1.62, x = 1; (b, f) ε=2.02, x = 2; (c, g) ε=2.54, x = 3; (d, h) ε=2.69, x = 4.

On the 2.5-mm-plane, the material does not undergo evident strain reversal due to the formation of the banded structure [23]. Therefore, the grain size only slightly increases due to the strain rate decreases.

3.6. Annealing stage

As shown in Fig. 3, the material enters the annealing region immediately after the completion of the material flow. Nevertheless, on the 0.5-mm-plane, the material is directly cooled by the rapid cooling, and thus, the annealing is negligible. The final microstructure is similar to that at x = 4. On the 2.5-mm-plane, the material stops flowing immediately once it moved to the rear side of the probe. Due to the coverage of the shoulder, the liquid CO2 cannot cool this region immediately. Therefore, the material undergoes significant annealing. As shown in Table 1, the grain size and Σ3 TB fraction increase continuously with the increase in the distance from the keyhole centre, whereas the LAB fraction decreases gradually. Fig. 13 shows the grain structures (a1, b1) and the corresponding orientation distribution functions (a2, b2) before and after annealing. At x = 2 (the end of the deceleration stage), the grain size (1.49 μm) is slightly larger than that at the high-velocity flow stage (1.10 μm). Nevertheless, their grain structures are similar (Figs. 10(b) and 13 (a1)). The texture mainly consists of strong B/$\bar{B}$, weak A/$\bar{A}$, C simple shear texture components [43] and 45° rotated cube component. After annealing (x = 8), the grain size increases to 2.03 μm and many Σ3 TBs are observed. The texture changes into strong 45° rotated cube and weak B/$\bar{B}$ components. The former indicates that discontinuous static recrystallization occurs during the annealing. The latter slightly deviates from its ideal orientations, which indicates that continuous change in the grain orientations may occur during the annealing. It may arise from the continuous static recrystallization, which usually occurs during the annealing of low-SFE metals with a large strain deformation [44,45]. Though these phenomena are observed on the 2.5-mm-plane, it is believed that similar phenomena would occur on the 0.5-mm-plane if no rapid cooling is used.

Table 1   Variations in the statistic grain size, LABs fraction and Σ3 TBs fraction at the annealing stage on the 2.5-mm-plane.

x value (mm)2468
Grain size (μm)1.491.641.822.03
LABs fraction0.3830.1560.1090.092
Σ3 fraction0.0470.2430.3170.367

New window

Fig. 13.   Grain structures (a1, b1) and corresponding orientation distribution functions (a2, b2) at the annealing stage (2.5-mm-plane): (a1, 2) x = 2; (b1, 2) x = 8. (Note: the R-cube here denotes the 45° rotated cube).

3.7. Overview of the dynamic development of the grain structure

The development of the grain structure depends on the specific history of the temperature, strain and strain rate. During FSW, the material first undergoes a preheating stage. This stage is believed to be inherent for general FSW processes. At the preheating stage, static recovery, recrystallization, and grain growth may occur, depending on the base material state. For example, in this study, the 1/4H work-hardened pure copper underwent static recrystallization and grain growth during this period, as described in Section 3.2.

Upon entering the SZ, the temperature is generally high [1,2] and roughly homogeneous [15,[20], [21], [22]] due to the heat conduction and convection around the rotation flow zone. In this region, the strain and strain rate dominate the evolution of the grain structure. The history of the strain and strain rate during the material flow is complicated. As described in Section 3.1, although the accumulative true strain is high, the strain at the acceleration and deceleration stages has different directions, i.e. strain reversal occurs. Simultaneously, the strain rates at the two stages undergo changes from zero to peak values and again to zero. Previous studies have reported the effect of strain reversal on the evolution of the grain structure during various deformation processes [[46], [47], [48], [49], [50], [51]], such as an increase in the recrystallization temperature, coarsening of the recrystallized grains, reduction of the dislocation density and mean misorientation angle, and retardation of the grain refinement.

During FSW, when the strain reversal occurs, the grain size does not change proportionally with respect to the increase in strain. Instead, the strain rate becomes the dominant factor for the development of the grain structure. Due to the strain reversal, there exists a stress relaxation period during which the strain rate is very low, and consequently, the strain hardening rate is much lower than the recrystallization rate (or grain growth rate), and thus, the grain size increases. When the reversal strain and the strain rate are sufficiently high, the grains can be refined again, as described in Section 3.5. Thus, after the deformation stage, the material still retains a significant amount of strain energy.

The strain rate influences the grain size during the plastic flow via the Zener-Hollomon parameter (Z factor) only under the strain saturation condition. In this study, at the acceleration stage, because the strain does not yet reach a saturated state, the strain rate has no direct effect on the grain size. Therefore, the grains are refined almost proportionally with the increasing strain, as shown in Fig. 5. Only when the strain is high enough and no strain reversal occurs, the grain size is controlled by the Z factor. For example, the grain size on the 2.5-mm-plane from x = -1 to x = 1 nearly keeps constant, as mentioned in Section 3.4. This is because the strain rate here is nearly constant. This is easy to understand because the material here is clinging to the probe surface and the probe has a constant rotation speed.

Subsequently, the weld material undergoes remarkable annealing during the cooling stage. During the annealing, preferred grain growth [52,53], normal grain growth [54,55], and static recrystallization (shown by this study) may occur, depending on the material properties such as the crystal structure and SFE, and processing parameters.

4. Conclusions

A quasi in situ method is used to quantitatively investigate the dynamic grain structure evolution considering the material flow, strain, and strain rate during FSW. The temperature was also qualitatively considered. The following conclusions were drawn from the results of this study:

(1)Corresponding to the material flow features, the microstructural evolution during FSW mainly undergoes five stages in time order, i.e., the preheating, acceleration flow, high-velocity flow, deceleration flow, and annealing stages.

(2)At the preheating stage, discontinuous static recrystallization and grain growth occur for the work-hardened base material. At the acceleration stage, the grains are refined by plastic deformation, DDRX, annealing twinning during HAB migration, and slight CDRX with the increase in strain. At the deceleration stage, the grain structure in the shoulder-affected zone first coarsens and is thereafter refined again due to the strain reversal; in the probe-affected zone, only a slight grain growth occurs. Finally, the hot-deformed microstructure in the shoulder-affected zone is ‘frozen’ directly by liquid CO2 whereas that in the probe-affected zone undergoes significant annealing, and thus, the recrystallized microstructure is obtained.

Acknowledgments

This study was partly supported by the New Energy and Industrial Technology Development Organization (NEDO) under the “Innovation Structural Materials Project (Future Pioneering Projects)” and a Grant-in-Aid for Science Research from the Japan Society for the Promotion of Science. One of the authors, Xiaochao Liu, thanks the China Scholarship Council for providing a scholarship.

The authors have declared that no competing interests exist.


/