Effect of initial texture on dynamic recrystallization of AZ31 Mg alloy during hot rolling
Maoyin Wang,Renlong Xin,Bingshu Wang,Qing Liu
DOI: https://doi.org/10.1016/j.msea.2010.11.069
2011-01-01
Abstract:Research highlights ▶ The initiation and completion of DRX in TD plate were retarded. ▶ {1 0 −1 1}–{1 0 −1 2} double twin related DRX nucleation was dominated in ND plate. ▶ Grain boundary related DRX nucleation was dominated in TD plate. Abstract Wedge-shaped AZ31 plates with two kinds of initial textures were rolled at 573 K to investigate the effect of initial texture on dynamic recrystallization (DRX). The results indicated that the initiation and nucleation of DRX were closely related to the initial texture. The initiation and completion of DRX in the TD-plate were significantly retarded compared with that in the ND-plate. Twin related DRX nucleation was mainly observed in the ND-plate samples; while gain boundary related DRX nucleation was mainly observed in the TD-plate samples. The different DRX behavior between the TD- and ND-plates was attributed to the different deformation mechanism occurring before DRX initiation. For the ND-plate, dislocation glide was considered as the main deformation mechanism accompanied with {1 0 −1 1}–{1 0 −1 2} double twin, which led to the increment of a faster increasing stored energy within the grains. And {1 0 −1 1}–{1 0 −1 2} double twin was mainly found to be DRX nucleation site for the ND-plate. For the TD-plate, {1 0 −1 2} extension twin was the dominant deformation mechanism which resulted in a basal texture with the c -axis nearly parallel to ND. The stored energy caused by dislocation motion was relatively small in the TD-plate before a basal texture was formed, which was considered as the main reason of that DRX was retarded in the TD-plate compared with that in the ND-plate. Based on the difference in deformation mechanism and DRX mechanism caused by the different initial texture, the variation in grain size, micro-texture and misorientation angle distribution in the ND and TD plates were discussed. Keywords AZ31 Mg Hot rolling Dynamic recrystallization Initial texture Twining Dislocation glide 1 Introduction During recent years, the use of magnesium alloys has been increased as structural materials. Especially the use of magnesium alloys has been increased significantly in 3c products such as cellular phones and notebook computers. This increase of the use of Mg alloys demands a technology development which can produce high quality sheet products with a high efficiency and a low cost. So, the development of rolling technology for Mg alloys is one of the hot areas in the recent years [1–5] . It is well known that dynamic recrystallization (DRX) plays a very important role during thermo-mechanical process of metals and alloys. For different kinds of deformation processes, such as extrusion and rolling, DRX has a strong effect on mechanical behavior (or formability) of Mg alloys during deformation process. On the other hand, DRX during thermo-mechanical process also has a strong effect on the formation of texture and grain refinement that are key factors for determining the mechanical properties of Mg alloys. So, the understanding of DRX of Mg alloys may shed light on the development of the rolling technology of Mg alloys sheets. From the literatures [6–8] , it was found that both continuous dynamic recrystallization (CDRX) and dis-continuous dynamic recrystallization (DCDRX) mechanisms were usually used to explain the experimental observations of DRX in Mg alloys. A continuous DRX phenomenon [6] was described as a progressive increase in grain boundary misorientation and conversion of low angle boundaries into high angle boundaries. However, in addition to the two types of DRX mechanisms, the DRX related to mechanical twining was termed as twin-DRX, and the DRX related to the high-angle boundaries formation promoted by the operation of ( a + c ) slip was termed as low-temperature DRX because ( a + c ) slip usually occurs at a relatively low temperature. In general, it has been accepted that DRX is a thermo-activated process of energy release (or stress release) when stored energy reaches a certain level during thermo-mechanical process. For Mg alloys, where DRX nuclei occurs and how DRX grains grow are directly related to the amount and the state of stored energy determined by plastic deformation mechanisms, such as dislocation slip and mechanical twining. So, it can be expected that there should be a direct relationship between DRX mechanisms and deformation mechanisms. In order to establish relationship among DRX mechanisms, initial texture, temperature and strain rate, many studies have been performed [9–11] on deformation behaviors of Mg alloys using uni-axial compression tests. However, research on DRX of Mg alloys during rolling process was rarely reported. For Mg alloys with hcp structure, initial texture has a strong effect on deformation mechanisms during plastic deformation. So, it can be expected that initial texture will also have a strong effect on DRX of Mg alloys during rolling process. So, the main aim of this paper is to investigate the effect of initial texture on DRX of Mg alloys during rolling at different temperatures. During recent years, there have been several publications [2,8,9,12] aiming to understand how DRX mechanisms are related to deformation mechanism of Mg alloys during thermo-mechanical processes. However, the conclusions from different researchers were very much divergent. Some researchers claimed that there is no direct link between deformation mechanism and DRX mechanism [2] , while others stated that deformation mechanisms have a strong effect on DRX mechanism [8,9,13] . So, one aim of the present paper is to clarify the issue of the relationship between deformation mechanism and DRX mechanism. 2 Materials and experimental procedures The material used in this study was a commercial Mg–3Al–1Zn hot-rolled plate. The as-received Mg alloy plate had a strong texture with the basal planes {0 0 0 1} aligned almost parallel to plate surface (see Fig. 1 ). Wedge-shaped plates, with a thickness variation of 1–5.8 mm from one end to another were cut from this plate for further rolling in order to have the deformed samples with different reduction from just one pass rolling. The plates were cut to have two initial textures related to the rolling geometry: one was cut with wedge surface approximately perpendicular to normal direction (ND) and another was cut with wedge surface approximately perpendicular to transverse direction (TD) of the original hot-rolling plate (see Fig. 2 ). The plates were termed as ND plate and TD plate, respectively. Parallel lines with even-distance were marked on the surface of the wedge-shaped samples in order to measure the thickness reduction at different position along the rolling direction of the samples after rolling. Before rolling, the samples were heated at 573 K for 2 h. By using a rolling mill with a roller of 450 mm in diameter, the samples were rolled to ∼1 mm thickness in one pass with a rolling speed of 1.2 m/s at 573 K. Blower air was used to cool down the samples after the rolling. By cutting at different positions along rolling direction, the samples with different rolling reductions were chosen for further characterizations of microstructure and texture. Microstructures of the samples with different rolling reductions were studied by optical microscopy (OM) and scanning electron microscopy (SEM, FEI Nova 400 FEG-SEM). Macroscopic and micro-texture of the samples were measured by X-ray diffraction (XRD, Rigaku D/max-2500) and electron backscatter diffraction (Oxford HKL Channel-5) techniques, respectively. 3 Results After hot rolling the two plates with different initial texture orientations, microstructures of the rolled-samples with different rolling reductions were obtained. Fig. 3 a–f shows the microstructures of the samples from ND plate, and Fig. 4 a–f are the microstructures of the samples from TD plate. Dynamic recrystallization grains related to twining were observed in the 10% rolled ND plate ( Fig. 3 b); while no DRX grain was observed in the 10% rolled TD plate ( Fig. 4 b). That means that DRX was retarded in TD plate compared with ND plate. As can be seen from Figs. 3c and d and 4c and d , not only the DRX was retarded but also the completion of DRX was retarded for TD plate compared with ND plate although there was a continuous progress of DRX in both kinds of plates with increasing strain. After 40% rolling reduction, DRX was observed to be completed in both cases ( Figs. 3e and 4e ) and no so much grain size or structure changes were observed in the further deformed (50% rolling reduction) samples ( Figs. 3f and 4f ). The orientation image and boundary structure maps ( Figs. 5 and 6 ) of samples with different rolling reductions were plotted by the data collected using a HKL Channle-5 EBSD system in a fine step size of 0.5 μm. Orientation image maps illustrate the angle values between c -axis and ND of the sheet, while the boundary structure maps illustrate both twin boundaries and grain boundaries which have misorientation angle larger than 5°. Several types of twin boundaries were observed ( Fig. 5 a and b) in the 10% rolled ND plate sample. By counting the length of each type of twin boundaries within the area of the EBSD map scanned, the relative length fraction of different types of twin boundaries were calculated and shown in Table 1 . Among the different types of twin boundaries, the relative length fraction of {1 0 −1 2} extension twin and {1 0 −1 1}–{1 0 −1 2} double twin boundaries are 35% and 43%, respectively (see Table 1 ), while others are less than 5%. In Figs. 3b and 5b , twin related DRX grains with a grain size of several μm are observed, and most of them are only related to {1 0 −1 1}–{1 0 −1 2} double twin boundaries, and no DRX grain related to {1 0 −1 2} extension twin was observed. A heterogeneous DRX structure of the 20% rolled ND plate sample was shown in Fig. 5 c and d. Extensive DRX occurred in some regions such as the upper and lower parts of Fig. 5 c and d, while deformed structure was observed at the middle part. The orientation difference in DRX and deformed structure regions are shown in Fig. 5 c. The DRX in the region with orientation of the c -axis away from ND was retarded compared with the regions with orientation of c -axis close to ND. With increasing strain, there was a progress of the completion of DRX observed in the 30% ( Fig. 5 e and f) and the 40% ( Fig. 5 g and h) rolled samples. It can be noted that the regions with orientation of c -axis closer to ND has a finer grain size compared with others. {1 0 −1 2} Extension twin boundary was observed in almost all the grains of the 10% rolled TD plate sample (see Fig. 6 a and b). The relative length fraction of the {1 0 −1 2} extension twin boundaries is more than 90% compared with other types of twin boundaries (see Table 1 ). Due to {1 0 −1 2} extension twining, the orientation of some of the grains was changed to the orientation with c -axis close to ND. No twin boundary was observed in several grains. Some of them may have the original orientation that did not favor any twining, while some of them may have the whole grain be twining. Similar structure with {1 0 −1 2} extension twin boundary was also observed in the 20% rolled TD plate sample ( Fig. 6 c and d). From both the 10% and 20% rolled samples, it was noted that in addition to {1 0 −1 2} extension twin boundaries, other types of twin boundaries, such as {1 0 −1 1}compression twin boundary and {1 0 −1 1}–{1 0 −1 2} double twin boundary, were observed in some of the grains. The DRX grains were observed to nucleate along the original grain boundaries in the 30% rolled TD plate sample (see Fig. 6 e and f) while no {1 0 −1 2} extension twin related DRX was observed. From Fig. 6 e and f, it can be noted that DRX grains have the orientation of c -axis closer to ND compared to the orientation of the regions with deformed structure. A completion DRX structure in the 40% rolled TD plate sample was shown in Fig. 6 g and h. From the color distribution showing the orientation in Fig. 6 g, the orientation of DRX grains is distributed very diffusely. Histograms of the misorientation angles across different boundaries of the rolled samples are shown in Figs. 7 and 8 . In order to have a detailed understanding on the characters of different types of boundaries, the misorientation rotation axes from three angle ranges are plotted in inverse pole figures ( Figs. 7 and 8 ). In all the rolled ND plate samples, there is a low angle peak distribution which is usually related to dislocation boundaries ( Fig. 7 ). The length of the low angle grain boundaries (2–4°, LAGB) per area in the ND and TD samples was listed in Table 2 . As can be seen, it increased from 0.08 in the 10% rolled sample to 0.16 in the 20% rolled sample, and then decreased to 0.10 in the 30% rolled sample and 0.05 in the 40% rolled sample. For the 10% rolled sample, in addition to the low angle peak distribution, the distribution peaks around 40° and 90° (see Fig. 7 a) are considered to be related to {1 0 −1 1}–{1 0 −1 2} double twin and {1 0 −1 2} extension twin boundaries, respectively. With increasing stain, the distribution peak around 30° appeared in the 20%, 30% and 40% rolled samples (see Fig. 7 b–d). The rotation axes of the 30° boundaries of the 20% rolled sample are mainly concentrated around 〈0 0 0 1〉 ( Fig. 7 b), while the concentration of the rotation axes became spreading with the development of DRX. After the completion of DRX in the ND plate (see Fig. 7 d), the boundary misorientation angle distribution for the recrystallization grains are mainly within the range of 10–60° with a peak distribution around 30°. For the rolled TD plate samples, the length of LAGB per area increased from 0.07 in the 10% rolled sample to about 0.22 in the 30% rolled samples, and then decreased sharply to about 0.04 in the 40% rolled sample. For the 10% rolled TD plate sample, there is a high distribution peak near 90° which is related to {1 0 −1 2} extension twin boundaries (see Fig. 8 a). With increasing strain, the relative frequency of the near 90° boundaries decreased and the distribution peak around 30° started to appear from the 30% rolled sample (see Fig. 8 c) and became to be more pronounced in the 40% rolled sample (see Fig. 8 d). The rotation axes of the 30° boundaries in both the 30% and 40% rolled samples are relatively concentrated around 〈0 0 0 1〉 ( Fig. 8 c and d). After the completion of DRX for TD plate (see Fig. 8 d), the boundary misorientation angle distribution for recrystallization grains are spreading within a wide range of 10–90° with a peak distribution around 30°. Based on EBSD measurements, the {0 0 0 2} and {1 0 −1 0} pole figures of the rolled samples to different strain from ND plate and TD plate are shown in Figs. 9 and 10 , respectively. For ND plate, there was no significant change of the basal texture during rolling deformation although the intensity of the texture became weaker with increasing strain (see Fig. 9 ). However, for TD plate, a significant texture change from {0 0 0 2} poles concentration around TD to those around ND was observed during rolling deformation. After the completion of DRX in the 40% and 50% rolled samples, a texture of {0 0 0 2} poles distribution between TD to ND was formed. 4 Discussions 4.1 Deformation mechanism and initiation of DRX From the experimental results shown above, there are two kinds of DRX nucleation, one is the twin related DRX nucleation observed mainly on the samples from ND plate (see Figs. 3b and 5c ), while another is the gain boundary related DRX nucleation observed mainly on the samples from TD plate (see Figs. 4d and 6e ). There are mainly two types of twining occurred on the samples from ND plates, one is {1 0 −1 2} extension twining and another is {1 0 −1 1}–{1 0 −1 2} double twining. From Fig. 5 c, it can be noted that twin related DRX nucleation was occurred mainly related to {1 0 −1 1}–{1 0 −1 2} double twining. The fine twin related DRX grains have {1 0 −1 1}–{1 0 −1 2} double twining orientation relationship with matrix. The twin related DRX may be related to the slip localizations due to a big orientation difference between two sides of twin boundary. However, for the TD plates samples, no obvious twin related DRX nucleation was observed although there were many {1 0 −1 2} extension twin boundaries observed in the samples. To be contrast, the nucleation of DRX on the TD plates samples was mainly related to the original grain boundaries. For the grain boundary related DRX, a relative early study on dynamic recrystallization of Mg alloy [14] under compression with a very low strain rate of 10 −5 s −1 showed that fine DRX grains were formed along the original grain boundary due to the localization slip of c + a dislocations near boundary. The difference of the DRX nucleation of the two types of plates studied in the present paper indicates that DRX mechanisms are closely related to deformation mechanisms which are very much dependent on the initial texture. In addition to the difference in nucleation between the two types of plates, another difference is DRX in TD plates was delayed, compared with ND plates. Similar results were found from previous work [11] on channel die compression tests of the samples with different initial texture. It was found that DRX was retarded in the samples with c -axis constrain compared with other orientations and it was explained that the microstructure requirements for the nucleation of dynamic recrytallization were not met readily under the predominance of prismatic slip which was encouraged by the constraining of c -axis. In the present work, the c -axis in the TD plate samples is parallel to the transverse direction of hot rolling, which is nearly the same as that in the channel die compressed samples with c -axis constrain. Recently, the samples prepared by equal channel angular pressing (ECAP) and hot rolling were used to study the influence of the initial texture on dynamic recrystallization and deformation mechanisms in AZ31 Mg alloys during tensile test [2] . It was found that the initial texture has a great influence on the amount of strain necessary to obtain DRX at moderate temperature. The deformation delay for the start of DRX of the ECAP sample was attributed to the strain necessary to develop a basal fiber with 〈0 1 −1 0〉 parallel to the tensile axis. Although it was assumed that the fiber texture stimulates DRX through the enhancement of multiple-slip, the authors still concluded that the influence of texture on DRX is not related to changes in the deformation mechanisms [2] . Comparing the present results with the results obtained from previous works [2,11,15] , it may be concluded that the strain at which DRX starts to occur is directly related to the strain at which the grain orientations have been changed to be close to basal texture. The mechanism of DRX in Mg alloys has also been analyzed by McQueen and Konopleva [16] . They proposed that the twinning takes place at low strains to reorient grains not suited for slip. As strain reaches a sufficiently high level, DRX sets in where high misorientations have been created by accumulation of dislocations, i.e. where slip has occurred on several slip system (near grain boundaries and twins). The new fine grains form a mantle (or a “necklace”) along grain boundaries and deform more easily than grain core, thus repeatedly undergoing recrystallization. From the experimental results above, the changes of the length of LAGB per area from low strain to high strain are different between the two kinds of plates. In fact, the length of LAGB per area should mainly be related to the amount of the dislocation boundaries accumulated within grains, which reflects the amount of the dislocation activity and the stored energy in the deformed metals. The maximum value of the length of LAGB per area is in the range of 0.16–0.22 which is nearly the same for both kinds of plates. However, the strain at which this value reaches a peak, or the strain at which this value begins to drop down is different for the two kinds of plates. For ND plate, the length of LAGB per area is 0.08 at the low strain sample (10% rolled sample), and then reached to a maximum value (0.16) at the 20% rolled sample. However, for TD plate, the length of LAGB per area is 0.07 at the low strain sample (10% rolled sample), and then increased to a maximum value (0.22) at the 30% rolled sample. The strains at which the maximum value reached for the two kinds of plates are corresponding to the strains at which DRX was observed. After the start of DRX, the length of LAGB per area dropped down rapidly. This difference of the length of LAGB per area in the two kinds of plates could be explained as that more dislocation activities occurred in ND plate at the low strain and the stored energy reached to be enough for initiating DRX at a lower strain compared with TD plate due to an un-favorable initial texture for twining. However, for TD plate, extension twining is the main deformation mode at low strain samples; the dislocation slips could only be dominant after the grain orientations were changed to be basal texture by twining. So, it will take more strain for TD plate to let the grains have enough stored energy for initiating DRX. The initial textures of the two plates in the present work can be considered as two extreme cases for comparison study, which may contribute to a better understanding of DRX mechanism in Mg alloys. In fact, in a recent study [15] on hot rolling of a coarse grained Mg–Al alloy under cast state, it was found an uneven rate of recrystallization at different grains which could be attributed to the different deformation mechanism occurred in the grains with different orientations. This uneven rate of DRX at different grains was considered to be the main reason [15] for the inhomogeneous microstructure formed during hot rolling. Although it was claimed [15] that the grains with basal texture are relatively stable and insensitive to further DRX, it was found that the initiation of DRX was related to the formation of the stable basal texture. In a polycrystalline Mg alloy, the grains can be classified into two groups, one with orientations near basal texture, and another with orientations away from basal texture. From the experimental results and discussions above, the grains with basal texture tended to start DRX earlier than those with non-basal texture. This can be attributed to the enhanced non-basal slip in the grains with basal texture. For the grains with orientations of non-basal texture, tensile-twining was usually formed to change grain orientation to basal texture before the initiation of DRX. 4.2 DRX grain orientation, grain size and mechanism There have been several studies [13,15,17] on the orientation (or texture) of DRX grains during hot deformation. In a study [13] on dynamic recrystallization of pure Mg and Mg–Y alloys during channel die compression, it was found that the texture of DRX grains followed closely that of the parent grains. From a study on large-strain hot rolling of Mg AZ61 [17] with a strong initial basal texture, the significant decrease of basal component during the first pass rolling was attributed to the operation of rotational dynamic recrystallization (RDX). In a recent study [15] on hot rolling of a coarse grained Mg–Al alloy under cast state, it was shown that DRX grains were formed in the twinned regions and then grew into the parent matrix by strain induced grain boundary migration. From the above experimental results, the DRX grains were formed in the twinned regions in the rolled ND plate samples ( Fig. 3 c) which has a strong initial basal texture. The growth of those DRX grains formed in the twinned regions into the parent matrix may also be used to explain the decrease of basal component in ND plate with increasing rolling strain ( Fig. 9 ). Although there is no obvious correlation between {1 0 −1 2} extension twin boundary and DRX nuclei in the deformed TD plate samples, from the experimental results above, it was noted that before the completion of DRX in TD plate with a strong initial texture of 90° from basal texture, the DRX grains have orientations closer to basal texture compared to those of deformed regions (see Fig. 6 e). This was further confirmed that DRX intended to occur within the grains after grain orientation has been changed to be close to basal texture by {1 0 −1 2} extension twinning. So, it may be expected that the DRX grains of the deformed TD plate samples should have similar orientations as those of ND plate due to the fact that DRX grains were formed within the regions which had been re-orientated to basal texture. However, since DRX usually starts within a grain before the whole grain has been twinned, DRX nuclei may also occur within the matrix regions which have the orientation of 90° from basal texture (the grains with blue color in Fig. 6 g) (for interpretation of the references to color in this sentence, the reader is referred to the web version of the article). So, compared with ND palate, there is a much wider orientation distribution for the DRX grains (see Fig. 8 d and e) in the deformed TD plate samples. It was also noted that there is a much wider boundary misorientation angle distribution (10–90° shown in Fig. 8 e) in the deformed TD plate samples compared with that (10–60° shown in Fig. 7 e) of the ND plate due to the different DRX grain orientations in the two kinds of plates in the present study. In fact, the orientation changes of different grains in the samples during hot rolling are much more complicated because it is a dynamic process in which dislocation slip or twining may occur at the new formed DRX grains just after the formation of them. The further slip or twinning will, of course, contribute to the orientation determination of the samples after hot rolling. So, the final structure and texture after hot rolling are the results of a balance between deformation process (slip and twinning) and the DRX process. Several studies on the amount and the grain size of DRX were reported [7,18] . It was found the amount and the size of dynamically recrystallized grains were increased as Zener–Hollomon parameter decreased. The amount of dynamically recrystallized grain was observed [7] to increase with strain in a sigmoidal form, while the related grain size was remained constant as strain increased. From a study on the effect of initial grain size on dynamic recrystallization of AZ31 Mg alloy [18] , it was found that there was no initial grain size dependence on DRX grain size when DRX proceeded to some degree, because the DRX grain size became constant for a given Z-value. That means the DRX grain size of Mg alloys is mainly dependent on the Z -value of the deformation process. The final DRX grain size of the ND samples and TD samples is very close in the present study indicating that the initial texture did not have significant effect on DRX grain size. This also confirmed that the DRX grain size of Mg alloys was mainly dependent on the Z -value of the deformation process. Hot torsion tests [8,16,19] have been used to study dynamic recovery and recrystallization of pure Mg and Mg alloys. Four different types of DRX mechanisms were proposed and it was found that DRX mechanisms are closely related with deformation mechanisms. The DRX related to mechanical twining was termed as twin-DRX. Based on the observation of the new fine grains which formed a mantle (or “necklace”) along original grain boundary, the model of “necklace” DRX was proposed [9,19] . Both types of DRX mechanisms were observed in the present work, and there is a strong effect of the initial texture on DRX mechanisms. The twin-DRX mechanism was observed mainly in ND plate with an initial basal texture, while the necklace DRX mechanism along original grain boundary was mainly observed in TD plates which have the orientation which favors to extension twining. In addition to the initial texture, it is well known that deformation mechanism will change with changing operating temperature. So it can be expected that DRX mechanism will be different at different operating temperature due to different deformation mechanism. A detailed work on the correlation of plastic deformation and dynamic recrystallization in Mg alloy ZK 60 was made by Galiyev et al. [9] . Different kinds of DRX mechanisms were proposed, and the mechanisms of DRX depended on the operating deformation mechanisms which changed with temperature. At temperatures below 473 K, twinning, basal slip and ( a + c ) dislocation slip were found to operate. The operation of ( a + c ) slip promoted the formation of high-angle boundaries and low-temperature DRX. In the intermediate temperature range (473–523 K), continuous DRX (CDRX) was observed and associated with extensive cross slip. At temperatures ranging from 573 to 723 K both bulging of original grain boundaries and sub-grain growth were the operating DRX mechanisms and controlled by dislocation climb. How the operating temperature affects deformation mechanism, and further affects DRX mechanism of both kinds of the plates in the present work needs to be further investigated. 5 Conclusions Wedge-shaped AZ31 plates with two kinds of initial textures, one with c -axes nearly parallel to normal direction (called ND plate) and another with c -axes nearly parallel to the transverse direction (called TD plate), were rolled at 573 K to investigate the effect of initial texture on dynamic recrystallization (DRX). It was found that the DRX happened during hot rolling was closely related to the initial texture. Based on the experimental results and discussions in the present work, the conclusions can be drawn as following: 1) Both {1 0 −1 2} extension twin and {1 0 −1 1}–{1 0 −1 2} double twin were observed to be very active in the rolled ND plates samples, while only {1 0 −1 2} extension twin was the dominant one in the rolled TD plates samples. 2) The initiation and completion of DRX in TD plate were significantly retarded compared with that in ND plate. {1 0 −1 1}–{1 0 −1 2} double twin related DRX nucleation was mainly observed in ND-plate samples; while gain boundary related DRX nucleation were mainly observed in TD-plate samples. 3) For ND plate, dislocation glide was considered as the main deformation mechanism which resulted in a faster increasing of the stored energy within the grains. For TD-plate, {1 0 −1 2} extension twin was a dominant deformation mechanism which resulted in a grain texture change to be basal texture with c-axis nearly parallel to ND. 4) Different DRX behavior on the rolled samples from the two kinds of plates was attributed to the different deformation mechanism occurring before the initiation of DRX on the plates with different initial texture. Lower energy was stored in TD-plate compared to ND-plate due to lesser dislocation glide occurred before a basal texture was formed, which was considered as the main reason of DRX was retarded in TD-plates. 5) The different deformation mechanism and DRX behavior caused by the different initial texture in ND and TD plates resulted in the different grain size, texture and misorientation angle distribution in the two kinds of rolled samples. Acknowledgements This project was financially supported by the National Basic Research Program of China (“973” Project) (Grant No. 2007CB613703 ) and the Natural Science Foundation of China (Grant No. 50890172 ). References [1] H. Xinsheng K. Suzuki A. Watazu I. Shigematsu N. Saito Mater. Sci. Eng. A 2008 214 220 [2] J.A. del Valle O.A. Ruano Mater. Sci. Eng. A 487 2008 473 480 [3] T. Al-Samman G. Gottstein Scripta Mater. 59 2008 760 763 [4] A. Styczynski C. Hartig J. Bohlen D. Letzig Scripta Mater. 50 2004 943 947 [5] S. Yi I. Schestakow S. Zaefferer Mater. Sci. Eng. A 516 2009 58 64 [6] J.C. Tan M.J. Tan Mater. Sci. Eng. A 339 2003 124 132 [7] S.M. Fatemi-Varzaneh A. Zarei-Hanzaki H. Beladi Mater. Sci. Eng. A 456 2007 52 57 [8] O. Sitdikov R. Kaibyshev Mater. Trans. 42 2001 1928 1937 [9] A. Galiyev R. Kaibyshev G. Gottstein Acta Mater. 49 2001 1199 1207 [10] Y. Zhang X. Zeng C. Lu W. Ding Mater. Sci. Eng. A 428 2006 91 97 [11] M.R. Barnett J. Light Met. 1 2001 167 177 [12] X.Y. Yang Z.S. Ji H. Miura T. Sakai Trans. Nonferrous Met. Soc. 19 2009 55 60 [13] R. Cottam J. Robson G. Lorimer B. Davis Mater. Sci. Eng. A 485 2008 375 382 [14] S.E. Ion F.J. Humphreys S.H. White Acta Metall. 30 1982 1909 1919 [15] Q.L. Jin S.Y. Shim S.G. Lim Scripta Mater. 55 2006 843 846 [16] M.M. Myshlyaev H.J. McQueen A. Mwembela E. Konopleva Mater. Sci. Eng. A 337 2002 121 133 [17] J.A. del Valle M.T. Pérez-Prado O.A. Ruano Mater. Sci. Eng. A 355 2003 68 78 [18] Y. Takigawa M. Honda T. Uesugi K. Higashi Mater. Trans. 49 2008 1979 1982 [19] S. Spigarelli M. El Mehtedi M. Cabibbo E. Evangelista J. Kaneko A. Jager V. Gartnerova Mater. Sci. Eng. A 462 2007 197 201