Microstructural evolution of pre-twinned Mg alloy with annealing temperature and underlying boundary migration mechanism

Ye Jin Kim, Jong Un Lee, Gyo Myeong Lee, Sung Hyuk Park

School of Materials Science and Engineering, Kyungpook National University, Daegu 41566, Republic of Korea

Received 13 July 2022; received in revised form 2 October 2022; accepted 31 October 2022

Available online 7 December 2022

Abstract This study investigates the variations in the microstructural characteristics of a pre-twinned Mg alloy with the temperature of the subsequent annealing treatment. To this end, a rolled AZ31 alloy is compressed to 3% plastic strain along the rolling direction (RD) to activate {10-12}twinning and is subsequently annealed at 200, 250, 300, 350, and 400 °C. Numerous {10-12} twins are formed throughout the compressed material, leading to the formation of a RD-oriented texture. At an annealing temperature of 200 °C, no microstructural variations occur during annealing. As the annealing temperature increases from 250 to 400 °C, the residual strain energy and remaining twin boundaries of the annealed material decrease owing to the promoted static recovery and the increased area fraction of twin-free grown grains. Consequently,an increase in the annealing temperature results in a gradual microstructural transition from a fully twinned grain structure to a completely twin-free grain structure. The microstructural evolution during annealing is predominantly governed by the movement of high-angle grain boundaries via a strain-induced boundary migration mechanism, and a few twin boundaries migrate above 350 °C because of their lower boundary energy. The boundary migration behavior and resultant microstructural evolution are discussed in detail based on the variations in boundary mobility and driving force for boundary migration with annealing temperature.

Keywords: Rolled Mg alloy; {10-12} twin; Grain growth; Boundary migration; Annealing.

1. Introduction

Mg alloys have a lower density and higher specific strength than commonly used structural metals, such as steels, Al alloys, and Ti alloys; therefore, they are highly desirable for lightweight structural applications. Automobile components made of Mg alloys provide various advantages over those made of other structural metals, including increased fuel efficiency, enhanced driving performance, and reduced CO2gas emissions. However, the application of Mg alloys in the external parts of automobiles has been limited owing to their insufficient mechanical properties compared to other structural metals [1]. During metal-forming processes (for example, forging, extrusion, and rolling), casting defects are eliminated [2], grains are refined [3], and the basal texture is strengthened [4]. Consequently, the strength and ductility of wrought Mg alloys subjected to metal forming are considerably higher than those of cast Mg alloys with coarse grains and casting defects [2,3]. However, the strong basal texture of wrought Mg alloys, which develops during hot metal forming process, causes low formability and asymmetric yield behavior at room temperature (RT) [5,6]. Extensive studies have been conducted to overcome these shortcomings by changing the texture of wrought Mg alloys through the addition of alloying elements that cause texture weakening or tilting (for example, rare earth elements and Ca) [7–9] or the application of metal-forming processes that induce shear deformation to a material (for example, equal-channel angular pressing and different speed rolling) [10–12]. Deformation twinning is an important deformation mechanism in Mg alloys, especially at RT, because of the insufficient number of active slip systems.Although various deformation twinning modes are present in Mg alloys, {10-12} extension twinning is overwhelmingly dominant in terms of the frequency of occurrence and area fraction of twins because of its lower critical resolved shear stress and higher twin boundary mobility than those of other twinning modes [13]. The occurrence of {10-12} twinning leads to the formation of twin boundaries in the parent grain and a lattice rotation of 86.3° in the twinned region. Therefore, the microstructure and texture of wrought Mg alloys can be effectively modified by generating {10-12} twinning.

Several studies have been conducted to improve the mechanical properties of Mg alloys using{10-12}twinning characteristics [14–16]. For instance, Lee et al. [17] reported that when a rolled AZ31 alloy is compressed along several directions, multiple {10-12} twin variants are generated, and the strength of the compressed alloy improves by 45% owing to the twin-induced grain refinement. Furthermore, {10-12} twin boundaries can act as nucleation sites for recrystallized grains during hot metal forming. Park et al. [18] revealed that an extruded AZ31 alloy fabricated using a {10-12}-twinned billet has a larger area fraction of recrystallized grains and a higher tensile strength and elongation than those fabricated using an un-twinned billet. Moreover, the texture modification induced by {10-12} twinning can lead to a change in the predominant deformation mechanisms of wrought Mg alloys [19]. For example, when a rolled AZ31 alloy is compressed to introduce{10-12} twins, its high-cycle fatigue properties improve considerably because the twin-induced textural change causes a variation in the main strain accommodation mechanism during cyclic deformation [20]. However, it has been reported that,although a compressed Mg alloy with {10-12} twins has a more desirable texture for formability than its initial twin-free counterpart, the RT formability of the former is lower than that of the latter owing to the increased residual strain energy by numerous dislocations formed during compression [15].Accordingly, additional heat treatment is required to relieve the high residual strain energy of compressed Mg alloys to improve their formability. Lee et al. [15] performed a 3-point bending test on a rolled AZ31 sheet subjected to precompression (PC) and subsequent annealing (PCA). They found that the application of the PCA process significantly improves the bendability of the rolled AZ31 sheet (by approximately 41%)because of a modified basal texture through precompression and decreased residual strain energy by subsequent annealing. In addition to bendability, recent studies have demonstrated that the PCA process can improve various properties of wrought Mg alloys, such as tensile ductility [21], stretch formability [22], and damping properties [23].

PC, which is the first treatment of the PCA process, activates {10-12} twinning in a wrought Mg alloy and consequently forms a new twin texture in which thec-axis is aligned nearly parallel to the compression direction. Subsequent annealing, which is the second treatment of the PCA process, reduces the residual strain energy and the number of twin boundaries of the PC-treated alloy.Interestingly,the twin texture of the PC-treated alloy is nearly maintained even after annealing. The microstructure and texture of the PCA-treated alloys and their mechanical properties are highly dependent on the amount of PC. Our previous studies revealed that the microstructure [24,25], texture [24], and bending properties[26] of PCA-treated rolled AZ31 alloys vary considerably with the amount of PC because the area fraction of the formed{10-12} twins and the intensity of the twin texture increase with increasing PC amount. However, the amount of PC that can be applied to wrought Mg alloys is limited by the shape of the material and the capacity of compression equipment. The temperature of subsequent annealing is also a crucial variable in the PCA process, because the microstructural evolution of deformed materials strongly depends on the heat-treatment temperature. Furthermore, the annealing temperature is not limited by the size of shape of the material and is easier to control than the amount of PC.A previous study[27]reported that the limiting bending depth of a PCA-treated Mg-Al-Zn-Ca-Y alloy increases by approximately 65% as the annealing temperature increases from 300 to 450 °C. This result indicates that the annealing temperature substantially affects the mechanical properties of the PCA-treated Mg alloys. However, detailed analyses have not yet been conducted on the influence of annealing temperature on the microstructural evolution during the PCA process and its underlying mechanisms.Therefore, this study systematically investigates the variations in the microstructure, texture, and grain growth behavior of rolled AZ31 samples subjected to 3% PC at annealing temperatures of 200, 250, 300, 350, and 400 °C.

2. Experimental procedure

A hot-rolled AZ31 (Mg-3Al-1Zn-0.5Mn, wt.%) plate with a thickness of 20 mm was used in this study. The plate was homogenized at 400 °C for 10 h in a box furnace to remove residual stress and solute segregation. Hereafter, the homogenized plate is referred to as the as-rolled (AR) material. For the PC treatment, rectangular samples of 14 mm in length,12 mm in width, and 10 mm in thickness [rolling direction(RD) × transverse direction (TD) × normal direction (ND)]were machined from the AR material. The machined samples were subjected to compression along the RD to a plastic strain of 3.0% (a total strain of 4.9%) at RT using a Shimadzu AGS-100kNX universal testing machine to form {10-12} twins in the specimens. The compressed samples were annealed at 200, 250, 300, 350, and 400 °C for 1 h in a box furnace, followed by water-quenching. The compressed sample and subsequently annealed samples are hereafter referred to as the PC-treated material and PCA-treated materials, respectively, and the PCA-treated materials at 200, 250, 300,350, and 400 °C are referred to as 200A, 250A, 300A, 350A,and 400A, respectively, according to their annealing temperatures.

Electron backscatter diffraction (EBSD) measurements were performed to analyze the microstructural characteristics(e.g., grain size, texture, twin boundary, and residual strain energy) of the AR, PC-treated, and PCA-treated materials.All samples for EBSD measurements were prepared by progressive mechanical polishing with #200 to #2000 grit sandpapers, 3-μm and 1-μm diamond compound pastes, and a colloidal silica solution. EBSD scanning was conducted on the RD–TD plane at the mid-thickness of all materials using a Hikari EBSD camera installed in a field-emission scanning electron microscope(Hitachi SU-70).Scanning data were collected using the TexSEM Laboratories (TSL) data acquisition software with a step size of 2.0 μm, and EBSD data were analyzed using the TSL orientation imaging microscopy analysis software. To ensure the reliability of the analysis results,data with a confidence index less than 0.1 were excluded from the analyses. In this study, the average grain size represents the average size of crystallographic domains surrounded by boundaries with misorientation angles of larger than 15°(i.e., high-angle grain boundaries (HAGBs) and twin boundaries). The microhardnesses of deformed and grown grains of the PC- and PCA-treated materials were measured using a Vickers hardness tester. All samples for hardness tests were prepared by etching the polished samples in an acetic picral solution (10 mL acetic acid + 3.0 g picric acid + 10 mL distilled water + 100 mL ethanol). After selecting the hardness measurement positions using an optical microscope, the hardnesses of the deformed and grown grains were measured with a low load of 0.1 kg to form indentation marks inside the grains for minimizing the influence of grain boundaries.After 15 hardness measurements for each grain, 13 values,excluding the maximum and minimum values, were averaged to obtain the hardness of the material.

3. Results

3.1. Microstructural characteristics of AR and PC-treated materials

Fig. 1a and b shows the inverse pole figure (IPF) map and the (0001) and (10-10) pole figures of the AR material. The AR material has an equiaxed grain structure without twin boundaries, and its average grain size is 39 μm.Thec-axes of most grains are aligned nearly parallel to the ND, whereas theira-axes are randomly aligned on the RD–TD plane. That is, the AR material has the typical basal texture of commercial rolled Mg alloys [13].The average value of the angle (θ) between thec-axis and RD (that is, the loading direction of PC) is approximately 79°, and almost all grains (area fraction: 98%) have an angle between thec-axis and ND of>45° (Fig. 1c).Fig. 1d shows, for Mg alloys, theoretical variations in the Schmid factors (SFs) for basal slip and {10-12} twinning of a grain as a function of the angle between thec-axis and loading axis. In this figure,αdenotes the angle between thea-axis of the grain and loading axis projected onto the basal plane.Atθ=79°andα=0°,the SFs for basal slip and{10-12}twinning during PC of the AR material are 0.19 and 0.35,respectively, whereas atθ= 79° andα= 30°, they are 0.16 and 0.48, respectively (Fig. 1d). Therefore, the average SFs for basal slip and {10-12} twinning under the PC condition of the AR material are approximately 0.18 and 0.42, respectively. Theθof most grains of the AR material is larger than 45° (Fig. 1c), which indicates that almost all grains have positive SF values for {10-12} twinning under the PC condition(Fig. 1d). Therefore, {10-12} twins can be formed throughout the material during PC.

The boundary map of the PC-treated material is shown in Fig. 2a, which reveals that {10-12} twins are formed in most grains of this material; however, there are almost no intersections of the {10-12} twin boundaries (blue boundaries in Fig. 2a). The formation of abundant {10-12} twin boundaries cause a decrease in the effective grain size from 39 to 28 μm.The (0001)pole figure of this material reveals that during PC,the ND-oriented texture weakens and a new RD-oriented texture forms because of the lattice reorientation of 86.3° caused by {10-12} twinning (Fig. 2b). It can be observed from the SF changes of six {10-12} twin variants (Fig. 1d) that the twin variant with the smallestθhas the highest SF [28]. This can be verified from Fig.2c,which shows the crystallographic orientations of the twin variants activated in the two grains with differentαvalues marked in Fig. 2b (that is, Grains 1 and 2). Only one twin variant is activated in Grain 1, withθ= 84° andα= 20°, whereas, four twin variants are activated in Grain 2, withθ= 86° andα= 9° (Fig. 2c). As shown in the pole figures in Fig. 2c, the basal poles of the twin variants activated in Grains 1 and 2 are oriented around the RD regardless of the number of activated twin variants.Because almost all grains of the AR material haveθ> 45°and the basal poles of the twinned region are nearly aligned along the RD, we denote the newly formed texture due to twinning (that is, texture of the region withθ< 45°) as “RD texture” and the texture of the residual matrix as “ND texture”. The area fractions of the twinned region (i.e., region with the RD texture) and the residual matrix (i.e., region with the ND texture) of the PC-treated material are 28% and 72%,respectively.

Fig. 1. Microstructural characteristics of the AR material: (a) IPF map, (b) (0001) and (10–10) pole figures, and (c) distribution of grain orientations. (d)Variation in Schmid factors for basal slip and {10-12} twinning as a function of angle between the c-axis of the grain and loading axis. Davg denotes the average grain size.

Fig. 2. Microstructural characteristics of the PC-treated material: (a) boundary map, (b) IPF map and (0001) and (10–10) pole figures, and (c) IPF maps,crystallographic relationship, and (0001) pole orientations of the residual matrix (M) and twinned region (T) of Grains 1 and 2 marked in (b).

3.2. Microstructural characteristics of PCA-treated materials

The IPF and boundary maps of the PCA-treated materials are shown in Fig. 3. The microstructure of 200A is nearly the same as that of the PC-treated material, both of which characterize the equiaxed grain shape and presence of numerous twin boundaries. The average grain size of 200A (27 μm) is also similar to that of the PC-treated material (28 μm). These results indicate that annealing at 200 °C does not cause microstructural variations in the PC-treated material. However,annealing at temperatures above 250 °C leads to noticeable variations in the microstructure. As the annealing temperature increases from 250 to 350 °C, the average grain size of the PCA-treated material gradually increases from 38 to 76 μm owing to the promoted boundary migration. Moreover, the formation and growth of regions devoid of twin boundaries causes a decrease in the area fraction of twinned grains formed during PC. The length of the twin boundaries per square millimeter decreases gradually from 0.11 to 0.009 mm-1with increasing annealing temperature from 200 to 350 °C, and almost all twin boundaries disappear during annealing at 400 °C; consequently, 400A exhibits a nearly twin-free grain structure. Because the twin-boundary-free region is formed by grain growth, such regions are hereafter denoted as “grown regions”; additionally, the grains containing the grown region are denoted "grown grains". In contrast,the regions and grains that do not grow during annealing are denoted as “deformed regions” and “deformed grains,” respectively. Some grown grains include the deformed region owing to the partial growth of the grains. The PCA-treated materials with a partially grown grain structure (i.e., 250A,300A, and 350A) contain grown regions, grown grains, deformed regions,and deformed grains,as shown in Fig.3c.IPF maps of the grown grains and residual deformed grains of the PCA-treated materials are shown in Fig. 4. No grown grains are formed during annealing at 200°C.As the annealing temperature increases from 250 to 350 °C, the area fraction of the deformed grains decreases from 65% to 19%, whereas that of the grown grains increases from 35% to 81%. At an annealing temperature of 400 °C, all the deformed grains disappear; consequently, 400A consists of only grown grains, in contrast with 200A, which comprises only deformed grains.

Fig. 3. Microstructural characteristics of the PCA-treated materials: IPF and boundary maps of (a) 200A, (b) 250A, (c) 300A, (d) 350A, and (e) 400A. Ltwin denotes the average length of {10-12} twin boundaries per square millimeter.

Fig. 4. IPF maps of the (a–e) deformed grains and (f–j) grown grains of (a, f) 200A, (b, g) 250A, (c, h) 300A, (d, i) 350A, and (e, j) 400A. fD and fG denote the area fractions of the deformed and grown grains, respectively.

Fig. 5. Variations in (a) area fractions of the deformed and grown grains and (b) average grain size and {10-12} twin boundary length per square millimeter with annealing temperature.

Fig. 6. (a) GOS maps and average GOS values of the PC-treated and PCA-treated materials. (b) Average hardness values of the deformed and grown grains of the PC-treated and PCA-treated materials.

Fig. 5a shows the variations in the area fractions of the deformed and grown grains with annealing temperature. The area fraction of the deformed grains decreases and that of the grown grains increases almost linearly with increasing annealing temperature. This suggests that as the annealing temperature increases, the microstructure of the PCA-treated material changes from a deformed grain structure to a grown grain structure, and a complete microstructural transition occurs at an annealing temperature of 400 °C. The length of the twin boundaries decreases with increasing annealing temperature because of the disappearance of deformed grains containing twins (Fig. 5b). In addition, the PCA-treated materials have higher average grain sizes than the PC-treated material owing to grain growth during annealing. However, the average grain size of the PCA-treated material does not increase linearly with annealing temperature; it increases up to an annealing temperature of 350 °C and then decreases when the annealing temperature increases from 350 to 400 °C. This variation in average grain size is discussed in more detail in Section 4.2.1.

Fig. 6a shows the grain orientation spread (GOS) maps of the PCA-treated materials. Because the grown regions formed during annealing at and above 250 °C have low strain energies, they appear bluish in the GOS maps. As the annealing temperature increases,the area fraction of the deformed grains with a high strain energy decreases,whereas that of the grown grains with a low strain energy increases. Accordingly, the overall strain energy of the material decreases gradually with increasing annealing temperature, which is evident from the decrease in the average GOS value. The average GOS value of 200A is 1.16, which is similar to that of the PC-treated material (1.20), indicating that the strain energy accumulated during PC is not relieved by subsequent annealing at 200 °C.As the annealing temperature increases from 250 to 400 °C,the average GOS value decreases from 0.85 to 0.47. Fig. 6b shows the average hardness values of the deformed and grown grains of the PC- and PCA-treated materials. The hardness of the deformed grains of the PC-treated material is 62.5 HV,and it gradually decreases to 54.1 HV as the annealing temperature increases to 350 °C, which is attributed to the promoted dislocation annihilation by static recovery at higher temperatures. Because the grown grains have a lower strain energy than the deformed grains, the hardness of the former is lower than that of the latter regardless of the annealing temperature,as shown in Fig. 6b for the hardnesses of 250A, 300A, and 350A. Hence, an increase in the annealing temperature leads to both a decrease in the hardness of the deformed grains and an increase in the area fraction of the grown grains with low hardness. However, when the annealing temperature increases from 350 to 400 °C, the hardness of the grown grains slightly increases from 52.4 to 53.7 HV. Because the grown grains of 400A are smaller than those of 350A (Fig. 3d and e), the former grains have a higher hardness than the latter grains owing to a stronger grain-boundary hardening effect. Consequently,the overall hardness of the PCA-treated material decreases with increasing annealing temperature up to 350 °C and then it slightly increases when the annealing temperature increases from 350 to 400 °C.

4. Discussion

4.1. Boundary migration mechanism

Fig. 7. IPF map, KAM map, and point-to-point misorientation profile along black arrows marked in IPF maps: (a) 250A and (b) 400A-1 min.

Fig. 7a shows the EBSD results of the selected grain of 250A, which is annealed at the lowest temperature among the annealing temperatures that cause the microstructural change.The grain morphology changes from equiaxed to irregular during annealing at 250 °C owing to grain growth. In the KAM map of Fig. 7a, the dotted white line indicates the site of the grain boundary before annealing, and the red arrow indicates the direction along which the boundary migration occurs during annealing. The line profile of the point-to-point misorientation angle along the black arrow in the IPF map of Fig. 7a reveals that a low-angle grain boundary (LAGB) is present between the deformed and grown regions. Zhao et al.[29] reported that when a pre-twinned Mg alloy is annealed at 250 °C, the microstructure changes from a twinned grain structure to a completely twin-free grain structure via static recrystallization during annealing. However, in the present study, static recrystallization does not occur during annealing of the PC-treated material. In the study by Zhao et al.[29], the pre-twinned material is fabricated by subjecting 7%cold rolling and 2% subsequent compression. Because this cold-rolled and compressed material has a substantially high residual strain energy, vigorous static recrystallization occurs during subsequent annealing. In this study, the PC-treated material is fabricated by subjecting only 3% compression,and therefore, its residual strain energy is not sufficient to generate recrystallization during subsequent annealing. In addition to the grain boundary migration at 250 °C (Fig. 7a),to analyze the microstructural changes during annealing at a high temperature, the microstructure of the PCA-treated material annealed at 400 °C for a short duration of 1 min (denoted as 400A-1 min) is shown in Fig. 7b. Similar to that in 250A, grain growth occurs along the direction of the red arrow marked in the KAM map of Fig. 7b, and an LAGB is present between the deformed and grown grains. These results indicate that microstructural evolution during annealing in the temperature range of 250–400 °C occurs through strain-induced boundary migration (SIBM). According to the SIBM mechanism [30], when arbitrary grain A has a higher strain energy than adjacent grain B in a deformed material,grain boundary migration occurs from grain B to grain A during the heat treatment to reduce the overall strain energy of the material, as shown in Fig. 8. The region newly formed by grain boundary migration has a low strain energy, and an LAGB is formed at the site of the HAGB that exists before heat treatment. Although the annealing temperature increases from 250 to 400 °C, microstructural variations, including the formation of grown grains and the grain coarsening, occurs via SIBM owing to the difference in the residual strain energy of the deformed grains of the PC-treated material. The residual strain energy of a deformed grain indicates the elastic strain energy accumulated in the grain during deformation, which is proportional to the degree of lattice distortion induced by dislocations formed during plastic deformation. In the PC-treated material, the dislocation density and resulting lattice distortion of individual grains differ owing to different activities of slip and twinning in each grain during PC [31,32]. Accordingly, SIBM occurs by the difference in local residual strain energy in the temperature range of 250–400 °C. However, the degree of microstructural changes (for example, the area fraction of the grown grains and average grain size) varies with the annealing temperature. To analyze the annealing temperature dependence of the microstructural evolution, the velocity of grain boundary migration during heat treatment needs to be considered, which is given by [33]

wherevis the grain boundary velocity,mis the grain boundary mobility,Pis the driving force for boundary migration,m0is the pre-exponential factor,kis the Boltzmann constant,Tis the temperature, andGmis the free energy of the boundary migration. Eq. (1) indicates that the grain boundary velocity is proportional to both the grain boundary mobility and driving force for boundary migration. Grain boundary mobility increases exponentially with increasing temperature(Eq. (2)); hence, grain boundary velocity increases as the annealing temperature increases. Therefore, at a given driving force for boundary migration, the boundary migration and resultant microstructural variations during the annealing of the PC-treated material are more pronounced at higher annealing temperatures.

4.2. Driving force for boundary migration

Fig. 8. Schematic illustration of the strain-induced migration of HAGB during annealing after PC.

According to Eq.(1),both the boundary mobility and driving force for boundary migration affect the grain boundary velocity. It is known that the driving force for boundary migration is in turn affected by several factors: stored strain energy, grain boundary energy, elastic energy, magnetic field,chemical driving force,surface energy,and temperature gradient [33]. However, when a deformed material is annealed, the dominant driving force for boundary mobility is the sum of the stored strain energy and the grain boundary energy [32].Additionally, solute atoms can affect the driving force for the boundary migration of a solid-solution alloy to some extent[33]. Detailed analysis of the factors that affect the driving force is presented in this section.

4.2.1. Stored strain energy

The driving force for the boundary migration (P) generated by the Gibbs free energy is expressed by the following equation [33]:

wheredGis the difference in Gibbs free energy between adjacent grains, anddVis the unit volume. The change in the Gibbs free energy (ΔGs) due to dislocation strain is given by[34]

whereξis the elastic strain energy per unit length of the dislocation,ρis the dislocation density, andVmis the molar volume. The driving force for boundary migration increases with increasing difference in the Gibbs free energy between adjacent grains (Eq. (3)), and the Gibbs free energy increases with increasing stored strain energy (Eq. (4)). Accordingly, a higher difference between the stored strain energies of two neighboring grains causes a larger grain boundary migration owing to the greater driving force. In this study, because the PC amount is constant as 3%, the samples before subsequent annealing have the same stored strain energy. However, the strain energy of each grain in the PC-treated material differs because the slip and twinning activities in individual grains during PC vary depending on their crystallographic orientation and size [31].

For the same grain boundary mobility, the change in the microstructure depends on the local driving force. At low annealing temperatures,because the overall mobility of the grain boundaries is low (Eq. (2)), only some grains with a sufficiently large local driving force for migration can grow. Once a grown region is formed via SIBM, it continues to grow because of its low strain energy. However, only a few grains undergo excessive growth during annealing at low temperatures because of the low frequency of encounters with other grown grains with low strain energy. In contrast, at high annealing temperatures, a relatively large number of grain boundaries can move owing to high boundary mobility. Fig. 9 shows the size distributions of the deformed and grown grains of the PC- and PCA-treated materials. The PC-treated material and 200A consist of only deformed grains with a normal size distribution. In 250A, because of the excessive migration of only a few grain boundaries, the average size of the deformed grains is low (23 μm), whereas that of the grown grains is relatively high (65 μm) despite the low annealing temperature.When the annealing temperature increases to 300°C,the average size of the deformed grains becomes 24 μm, which is almost the same as that of 250A. However, the average size of the grown grains increases from 65 to 84 μm owing to the increased boundary mobility at the higher temperature. For 350A,the average sizes of the deformed and grown grains(28 and 88 μm, respectively) are similar to those of 300A. However, the average grain size of 350A (76 μm) is larger than that of 300A (63 μm) because of the higher area fraction of the grown grains in the former.Interestingly,the average grain size of 400A is 57 μm, which is smaller than those of 300A and 350A. The average sizes of the deformed grains of all the PC-and PCA-treated materials are similar(23–28 μm),which indicates that the deformed grains remain nearly unchanged during annealing irrespective of the annealing temperature.However, the average size of the grown grains increases significantly from 65 μm for 250A to 84 μm for 300A and increases slightly to 88 μm for 350A, but decreases drastically to 57 μm for 400A.

Fig. 9. Size distributions of the deformed and grown grains of (a) the PC-treated material, (b) 200A, (c) 250A, (d) 300A, (e) 350A, and (f) 400A. DD, DG,and Davg denote the average sizes of the deformed grains, grown grains, and all grains (deformed grains + grown grains), respectively.

400A, which is annealed at the highest temperature, has a twin-free equiaxed microstructure with high grain size homogeneity. When the grown grains are adjacent to each other,microstructural changes rarely occur because the boundaries shared by the grown grains have a low driving force for migration owing to the small difference in the strain energies of the adjacent grown grains. The locations of boundaries in which SIBM is activated during annealing can be considered as nucleation sites for boundary migration. As the annealing temperature increases, the number of the nucleation sites also increases. Hence, during annealing, a growing grain comes in contact more rapidly with another growing grain, which leads to a reduction in the degree of grain growth. The variation in microstructure of a deformed material during subsequent heat treatment occurs by competition between the nucleation and growth, and the nucleation and growth behaviors are considerably affected by heat treatment temperature [30]. Therefore, depending on the annealing temperature, the difference in the microstructural evolution behavior induces a difference in the grain size distribution of the PCA-treated material. The number density of the nucleation sites is relatively small in 250A; hence, only a few grown grains are in contact with each other, as shown in Fig. 4g. However, as the annealing temperature increases, the number density of the nucleation sites increases, and therefore, more grown grains come in contact with each other. Therefore, although the boundary mobility increases with increasing annealing temperature, additional grain growth is suppressed when the microstructural transition from a twinned grain structure to a grown grain structure occurs sufficiently during annealing. At a high annealing temperature of 400°C,the grain boundaries can move even with a small driving force for migration.Therefore,most grains grow uniformly during annealing at this temperature,which suppresses the degree of grain growth owing to early contact with each other. In other words, at 400 °C, a large number of the nucleation sites for boundary migration inhibit the growth of grown grains, which results in the formation of a relatively small size of grown grains due to the promoted nucleation and suppressed growth behaviors. Consequently,the rate of increase in the grown grain size gradually decreases when the annealing temperature increases to 350 °C,and finally, the grown grain size decreases when the annealing temperature increases to 400 °C. Therefore, the average grain size of 400A, which has a fully grown grain structure,is smaller than that of 300A, which has a partially grown grain structure.

4.2.2. Boundary energy

The boundary energy is also an important driving force(Pg) for boundary migration [33], which is represented via equation by [33]

Fig. 10. EBSD results of 400A-1min: (a) IPF map of entire measured region and (b, c) IPF and KAM maps of regions (b) A and (c) B marked by white rectangles in (a). The black and white arrows in (b) and (c) denote the direction of boundary migration and the migrated twin boundary, respectively.

whereγis the boundary energy, andRis the radius of curvature of the boundary. The boundary energy is an intrinsic property of the boundary, and a higher boundary energy leads to a higher boundary velocity (Eq. (5)). There are two types of boundaries in the PC-treated material: HAGB and {10-12}twin boundary (Fig. 2a). The boundary energy of HAGB is as high as 0.4–1.3 J·m-2[35–37], whereas that of {10-12}twin boundary is as low as approximately 0.1 J·m-2[38,39].Therefore, the driving force for twin boundary migration derived from the boundary energy is significantly lower than that of the HAGB. In a previous study [32], microstructural variations of a pre-twinned AZ31 alloy during annealing at 250 °C for 1 h were analyzed usingquasi in-situEBSD. The results revealed that during annealing, the HAGBs move actively, but the {10-12} twin boundaries do not move [32]. In the present study, most of the twin boundaries in the deformed grains do not move and maintain a straight shape during annealing at 250–350 °C. Fig. 10 shows the IPF and kernel average misorientation(KAM)maps of 400A-1 min.The twin boundaries in the deformed grains of 400A-1 min mostly have a straight shape, except for a small number of twin boundaries that are curved (Fig. 10a). The grown grain in region A of Fig. 10a comprises a residual matrix (M1) and a twin (T1), as shown in Fig. 10b. M1grows along the direction of the black arrows and consumes the surrounding regions. Furthermore, the newly formed M1region with low strain energy encroaches the adjacent T1region with high strain energy through twin boundary migration (the white arrow in the KAM map of Fig. 10b) despite its lower boundary energy. This is because the twin boundary has a high driving force for migration owing to the large difference in the strain energies of T1and M1and the high twin boundary mobility at this high temperature. Another case in which twin boundary migration occurs can be observed in region B of Fig. 10a. Fig. 10c shows that region B consists of two grains: each of which comprises a residual matrix and a twin (M2and T2and M3and T3), and they are in contact with each other. The M2-M3and T2–T3interfaces have low misorientation angles of approximately 2°M2grows along the direction of the black arrow and forms a grown grain with a low strain energy (M4), as shown in Fig. 10c. Meanwhile, T3grows along the direction of the yellow arrow and consumes T2(Fig. 10c). Consequently, an interface forms between M4and the grown T3region. This interface, in turn, gives rise to a curved twin boundary owing to the small difference in the misorientation angle between T2and T3. (the twin boundary indicated by the white arrow in the KAM map of Fig. 10c). When a matrix region with low strain energy adjoins a twinned region with high strain,twin boundary migration can occur to reduce the strain energy. This strain-induced twin boundary migration is similar to the HAGB migration via SIBM, which leads to the formation of curved twin boundaries during annealing. However,owing to the low boundary energy, curved twin boundaries can be formed at high annealing temperatures. The boundary maps in Fig. 3 show that a few curved twin boundaries,which are marked by the black arrows in the boundary maps,are present in only 350 A and 400 A that are annealed at high temperatures. This implies that twin boundary migration occurs to some extent during annealing at temperatures above 350 °C. However, its influence on the microstructure and texture of the PCA-treated material is insignificant because the microstructural evolution during annealing is predominantly governed by HAGB migration.

4.2.3. Solute segregation

A pre-twinned AZ31 alloy was subjected to annealing at 170 and 200 °C by Xin et al. [40]; they found the segregation of Al and Zn solute atoms at the {10-12} twin boundaries of the annealed alloy. In addition to solute segregation at twin boundaries, Pei et al. [41] reported that when an AZX310 alloy is heat-treated at 420 °C, the Al, Zn, and Ca solute atoms are segregated at HAGBs. Such twin and grain boundary segregations occur because solute atoms, which are zerodimensional lattice defects, diffuse into the two-dimensional boundary defects during heat treatment to reduce the Gibbs free energy of a material [42]. Solute segregation at a boundary decreases its mobility owing to the solute drag effect[33].The drag force (Pv) exerted by a solute on the grain boundary can be expressed as follows [33]:

wheren0is the number of lattice sites per unit boundary area,fis the force of attraction between the boundary and a foreign atom,cbis the solute concentration at the boundary,c0is the volume impurity concentration,Uis the interaction energy,kis the Boltzmann constant, andTis the temperature. The drag force is affected by the boundary properties, the total amount of dissolved solute,and temperature(Eq.(6)).If the boundary properties and amount of dissolved solutes remain unchanged,then the amount of solutes that are segregated at the boundaries decreases as the temperature increases. From Eq. (1),the boundary migration velocity, considering the solute drag effect, is expressed asv=mb· (P-Pv). As the annealing temperature decreases,Pvincreases and, consequently,the boundary migration velocity decreases. Therefore, both HAGB and twin boundary migrations during annealing can be suppressed owing to the solute drag effect, and this effect is more pronounced at lower annealing temperatures.

Fig. 11. Schematic illustrations showing microstructural evolution behavior during annealing at low and high temperatures after PC.

The microstructural evolution behaviors during annealing at low (≤300 °C) and high (≥350 °C) temperatures are shown schematically in Fig. 11. When the PC-treated material is annealed at low temperatures (≤300 °C), a relatively small number of HAGBs with a high driving force migrate owing to the low boundary mobility and high solute drag effect at these temperatures, and these migrated HAGBs form a new region with low strain energy (that is, the grown region). The grown region continues to grow through further migration of the HAGBs because of the large difference in the strain energies between the grown region and adjacent deformed regions. However, twin boundaries do not migrate at these temperatures owing to their low boundary energies.At high annealing temperatures (≥350 °C), most HAGBs,including those with a low driving force, start to migrate because of the enhanced boundary mobility induced by the high temperature. When a grown region comes into contact with a nearly grown region,boundary migration no longer occurs because of the small difference in the strain energies of the two grown grains.Consequently,the average migration distance of the HAGBs decreases at high annealing temperatures because the grown grains are homogeneously formed throughout the material. In summary, the nucleation and growth behaviors of grown grains through the migration of HAGBs vary with the annealing temperature, which results in differences in the microstructural characteristics after annealing, such as the size of the grown grains, amount of remaining twin boundaries,and area fraction of residual deformed grains.

4.3. Texture evolution behavior

Fig. 12. (0001) pole figures of the PCA-treated materials: (a) overall texture, (b) ND texture, and (c) RD texture.

Fig. 13. Characteristics of the twinned and matrix regions of the PC-treated material: (a) distributions of grain orientations, (b) SF distributions for basal slip under deformation along the RD, and (c) IPF and SF maps for a selected area and SFs for basal slip of matrices (M1, M2, M3, and M4) and {10-12} twins(T1, T2, T3, and T4). SFM and SFT denote the average SFs for basal slip of the matrix and twinned regions, respectively.

Fig. 12 shows the (0001) pole figures of the PCA-treated materials. As described in Section 3.1, the PC-treated material has two types of textures: RD texture, corresponding to the twinned region, and ND texture, corresponding to the residual matrix. Because the microstructure of the PC-treated material remains unchanged during annealing at 200 °C, the texture of 200 A is almost the same as that of the PC-treated material. As the annealing temperature increases from 250 to 400 °C, the texture gradually changes, as shown in Fig. 12.All the PCA-treated materials have both RD and ND textures,and their maximum texture intensities are similar (Fig. 12a).However, the ND and RD textures exhibit different evolution behaviors during annealing. The ND texture is dispersed around the ND after annealing, whereas the RD texture is more concentrated toward the RD after annealing (Fig. 12b and c). The different evolution behaviors of the ND and RD textures are caused by the difference in residual strain energy of the matrix region and twinned region during PC. In the PC-treated material, theθrange of the twinned region,which has the RD texture, is relatively close to 45° compared to that of the matrix region, which has the ND texture(Fig. 13a). This is because that when a rolled Mg material is compressed along the RD, a twinned region formed in a matrix region withθ= 90° can haveθranging from 0°to 30° [43]. Therefore, the average SF for basal slip of the twinned region (0.26) is higher than that of the matrix region(0.20)in the PC-treated material(Fig.13b).Furthermore,Wang and Agnew [44] reported that dislocations are transmuted from basal dislocations in {10-12} twins and the twins are hardened by the dislocations during further deformation because these dislocations are usually sessile. Consequently, the twinned region has a higher residual strain energy than the matrix region. As shown in Fig. 13c, the PC-treated material has three types of HAGBs:(i) a HAGB between a matrix region and adjacent matrix region (i.e., matrix–matrix boundary), (ii) a HAGB between a matrix region and adjacent twinned region (i.e., matrix–twin boundary), and (iii) a HAGB between a twinned region and adjacent twinned region (i.e., twin–twin boundary). The matrix region has a lower SF for basal slip than the twinned region, which is evident from the lower SFs of matrices M1,M2, M3, and M4(0.02–0.11) than those of {10-12} twins T1,T2, T3, and T4(0.15–0.47), as shown in Fig. 13c. Hence,when the matrix region is in contact with the twinned region,the matrix region grows by consuming the twinned region via the migration of the matrix–twin boundary. The matrix region with an off-basal texture in which basal poles are somewhat deviated from the ND can readily grow during annealing because of the presence of twinned region in contact with them;consequently, their growth leads to the dispersion of the ND texture after annealing. The twinned region can grow via the migration of the twin–twin boundary. When two twinned regions are in contact with each other, the twinned region with a smallerθgrows by consuming the other with a relatively higherθbecause the former has lower strain energy due to its lower SF for basal slip. The growth of the twinned region with a smallerθand the consumption of the twinned region with a higherθthrough the migration of twin–twin boundaries during annealing result in the concentration of the RD texture toward the RD after annealing.

5. Conclusion

This study investigates the variations in the microstructure of a pre-twinned rolled Mg alloy with annealing temperatures in the range of 200–400 °C. When the AR material is subjected to PC along the RD, abundant {10-12} twins are formed in most grains, which reduces the average grain size and changes the basal texture from an ND texture to an RD texture. During annealing at 200 °C after PC, the grain size and twinned structure of the PC-treated material remain unchanged and only its strain energy decreases by static recovery. As the annealing temperature increases from 250 to 400 °C, the area fraction of twin-free grown grains with a low strain energy increases gradually, resulting in a microstructural transition from a fully twinned grain structure to a completely twin-free grain structure. Microstructural evolution during annealing predominantly occurs through the migration of HAGBs via the SIBM mechanism.The degree of boundary migration increases with increasing annealing temperature because of the increase in boundary mobility and decrease in the solute drag effect. Consequently, the average grain size of the PCA-treated material increases as the annealing temperature increases from 200 to 350 °C. However,the average grain size decreases when the annealing temperature further increases to 400 °C because most grains grow uniformly during annealing at this temperature and the early contact of the grown grains suppresses the degree of grain growth. The migration of HAGBs with a high boundary energy occurs at annealing temperatures above 250 °C, whereas only a few {10-12} twin boundaries migrate above 350 °C owing to their lower boundary energy. The ND texture of the PC-treated material is dispersed around the ND after annealing, whereas its RD texture is more concentrated toward the RD after annealing. This study demonstrates that annealing temperatures between 200 and 400 °C have a significant effect on the grain size, residual strain energy, remaining twin boundaries, and texture of the PCA-treated material because of the different boundary migration behaviors during annealing. In addition, wrought Mg alloys with a twin-free equiaxed grain structure and a modified basal texture can be fabricated through annealing treatment above 400 °C after PC.

Data availability

The raw/processed data required to reproduce the findings of the present study cannot be shared at this time, as the data are part of an ongoing study.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have influenced the work reported in this paper.

CRediT authorship contribution statement

Ye Jin Kim: Methodology, Data curation, Formal analysis, Writing – original draft. Jong Un Lee: Investigation, Visualization. Gyo Myeong Lee: Investigation, Data curation.Sung Hyuk Park: Conceptualization, Writing – review &editing, Project administration, Supervision.

Acknowledgments

This research was supported by the National Research Foundation of Korea (NRF; Grant No. 2019R1A2C1085272)funded by the Ministry of Science, ICT, and Future Planning(MSIP, South Korea).