Towards Understanding Relationships between Tension Property and Twinning Boundaries in Magnesium Alloy

: Although pre-induced (cid:8) 10 12 (cid:9) twins could strengthen magnesium and its alloys, the origin of such a strengthening phenomenon remains questionable. This is because twins can simultaneously change the size of grains and the texture features of the initial material. In the present work, the effect of pre-induced (cid:8) 1012 (cid:9) twins on the tension property of an extruded magnesium alloy has been investigated through a combination of electron backscatter diffraction, transmission electron microscope, and mechanical tests. Samples with and without (cid:8) 1012 (cid:9) twinning boundaries, but possessing an almost identical texture characteristic, were prepared by pre-compression perpendicular to the extrusion direction. Subsequently, these pre-strained samples were tensioned along the extrusion direction. The results indicate that the pre-induced (cid:8) 1012 (cid:9) twinning boundaries can indeed enhance the tension strength of magnesium alloys, but only slightly. The effect is closely associated with the amount of pre-strain. Correspondingly, the possible mechanisms behind such phenomena are given and discussed.


Introduction
Magnesium and its alloys have obtained extensive attention for their application in the automotive and aerospace industries because of their outstanding properties, such as low density and high specific strength [1]. However, as a typical metal with a hexagonal close-packed (hcp) structure, in addition to dislocation slips, deformation twins also play an important role during plastic deformation, especially at low temperatures [2][3][4]. This leads to magnesium and its alloys possessing a strong, basal texture characteristic after rolling and extrusion processes, and exhibiting a remarkable tension/compression yield asymmetry, due to the polar feature of deformation twins [5][6][7]. Besides, magnesium and its alloys also exhibit a low yield strength at room temperature [1,[4][5][6][7][8]. These characteristics significantly restrict their processing, design, and applications, especially as load-bearing components. A number of previous publications have indicated that refining grain size is an effective procedure to reduce the disadvantages of magnesium alloys, and, correspondingly, several attempts have been made to produce magnesium alloys by refining grain size [9][10][11][12]. For example, by means of equal-channel angular pressing (ECAP), Li et al. [9] successfully obtained AZ60 magnesium alloy with grain size of~0.8 µm after eight passes and improved the mechanical property of magnesium alloy at a high strain rate.
Recently, some publications have suggested that pre-induced 1012 twins can also refine the grain size of initial magnesium alloys, resulting in the enhancement of their yield strength [13][14][15][16]. The authors suggested that 1012 twinning boundaries (TBs) may have a similar function to traditional high-angle grain boundaries that can strongly impede the movement of matrix dislocations [15][16][17]. As a consequence, matrix dislocations have to halt and be piled up in the front of 1012 TBs during plastic deformation, leading to the strengthening effect in magnesium alloys. By contrast, based on experimental observations, M.R. Barnett suggested that, compared to grain refined by TBs, reorientation of the c-axes by twins might play a more important role in the strengthening effect in magnesium alloys [18]. Knezevic et al. [19] further pointed out that the main contribution of 1012 twins to hardening came from texture hardening, namely that the initial grains with soft orientations were changed into new grains with hard orientations by 1012 twins. Similar conclusions were also reported by Wang et al. [20]. More recently, a series of works conducted by Chen et al. [21,22] even suggested that, during plastic deformation, 1012 TBs appeared to be an ineffective barrier for the matrix dislocation movement, indicating that the strength caused by 1012 TBs might be negligible. From those above studies, it is quite clear that pre-induced 1012 twins can significantly influence the mechanical property of magnesium alloys, although a consensus has not been reached about the causes of these effects.
For this purpose, in the present work, the influence of pre-induced 1012 twinning boundaries on the strength of extruded AZ31 samples was investigated by a combination of electron backscatter diffraction (EBSD), transmission electron microscope (TEM) and mechanical tests. The results show that 1012 TBs themselves can indeed enhance the strength of magnesium alloy, but only slightly. Accordingly, the possible mechanisms behind such interesting phenomena are discussed. These findings are expected to provide a new insight into the role of the 1012 twin in magnesium and other metals with hcp structure.

Initial Material and Deformation
It should be noted that for magnesium and its alloys, when 1012 TBs are introduced, the texture feature is usually changed. Thus, one intractable problem is how we can fabricate a magnesium alloy sample that contains 1012 TBs but still possesses the initial texture feature. In this case, the influence stemming from pre-induced TBs themselves can be evaluated conveniently. It is well known that extruded magnesium alloys usually possess a strong basal fiber texture, namely that the (0001) crystallographic direction of grains is largely perpendicular to the extrusion direction (ED), and randomly distributed around the ED [8]. On the other hand, as indicated in references [23,24], the 1012 twin prefers to rotate the lattices of initial grains into the external load direction. Accordingly, based on the features of extruded magnesium alloy and the 1012 twin, we may fabricate a special sample which satisfies the above requirements.
The material used in the current work was an extruded AZ31 (3 wt% Al-1 wt% Zn) magnesium alloy rod. As presented in Figure 1a, the as-received material displays a twinfree microstructure feature with an average grain size of ∼ 30 µm. The corresponding pole figures shown in Figure 1b indicate that the extruded rod presents a typical extrusion texture with (0001) poles largely perpendicular to the ED and randomly distributed around the ED. In order to obtain samples with pre-induced 1012 TBs, some rectangular parallelepiped samples with the dimensions of 100 mm × 5 mm × 10 mm (ED × ERD × ETD) were first cut from the center of the as-received material by electrical discharge machining (EDM), where the ERD and ETD are referred to as the radial direction and tangential direction of the extrusion rod, respectively. Then, one part of these samples was subjected to pre-compression along ERD at a strain rate of 1 × 10 −3 s −1 to a strain of~0.9% and~1.3%. Subsequently, all of these samples, including samples with and without pre-compression, were annealed at 200 • C for 6 h to remove the dislocations caused by pre-strain, while maintaining the twin structure [14]. Accordingly, those samples with and without pre-strain subjected to anneal are denoted as ERD0.9CA, ERD1.3CA and ERD0CA, respectively, in the succeeding paragraphs.

ED ERD ETD
× × ) were first cut from the center of the as-received material by electric discharge machining (EDM), where the ERD and ETD are referred to as the radial dire tion and tangential direction of the extrusion rod, respectively. Then, one part of the samples was subjected to pre-compression along ERD at a strain rate of 3 1 1 10 s − − × to strain of ~0.9% and ~1.3%. Subsequently, all of these samples, including samples with an without pre-compression, were annealed at 200 C° for 6 h to remove the dislocatio caused by pre-strain, while maintaining the twin structure [14]. Accordingly, those sam ples with and without pre-strain subjected to anneal are denoted as ERD0.9CA ERD1.3CA and ERD0CA, respectively, in the succeeding paragraphs. . In order to measure the accura strain, an extensometer (E98918, Epsilon) was mounted on every sample. Three sampl for each condition were stretched to a cracking point for a better understanding of th relationship between microstructure feature and macroscopy performance. Two tensio tests for each condition were disrupted at a strain of ~1.2%. Correspondingly, these sam ples are marked as ERD0.9CA-ED1.2T, ERD1.3CA-ED1.2T and ERD0CA-ED1.2T, respe tively.

EBSD and TEM Examinations
The microstructural characteristics of the ERD-ETD cross section of these deforme samples were examined by EBSD. Before EBSD examination, the investigated surfac were ground mechanically with green SiC papers from 500# to 4000#, followed by ele tropolishing (voltage: 15 V, electric current: 0.15 A, temperature: −30°, dwell time: 150 using the commercial AC2 solution. The EBSD investigations were performed via th HKL Channel 5 system equipped in FEI Nova 400 FEG-SEM. The step size used in th In order to investigate the effect of pre-induced 1012 TBs on the subsequent tension property, five dog-bone shaped tension samples with dimensions of the rectangular gage section of 50 mm × 4 mm × 5 mm (ED × ERD × ETD ) were cut from ERD0.9CA, ERD1.3CA and ERD0CA by EDM technology, respectively. Then, tension mechanical tests along ED were performed at a strain rate of 1 × 10 −3 s −1 . In order to measure the accurate strain, an extensometer (E98918, Epsilon) was mounted on every sample. Three samples for each condition were stretched to a cracking point for a better understanding of the relationship between microstructure feature and macroscopy performance. Two tension tests for each condition were disrupted at a strain of~1.2%. Correspondingly, these samples are marked as ERD0.9CA-ED1.2T, ERD1.3CA-ED1.2T and ERD0CA-ED1.2T, respectively.

EBSD and TEM Examinations
The microstructural characteristics of the ERD-ETD cross section of these deformed samples were examined by EBSD. Before EBSD examination, the investigated surfaces were ground mechanically with green SiC papers from 500# to 4000#, followed by electropolishing (voltage: 15 V, electric current: 0.15 A, temperature: −30 • , dwell time: 150 s) using the commercial AC2 solution. The EBSD investigations were performed via the HKL Channel 5 system equipped in FEI Nova 400 FEG-SEM. The step size used in the current EBSD examinations was 1 µm. For TEM and high-resolution TEM (HRTEM) observations, the samples were cut for the center of ERD1.3CA, and then thinned by grinding to 50 µm, followed by the low-temperature ion thinning technique. Subsequently, TEM and HRTEM were performed by a FEI Tecnai F20-G 2 electron microscope with a voltage of 200 kV.

Mechanical Properties
Stress-strain curves under tension along ED for the ERD0.9CA, ERD1.3CA and ERD0CA are given in Figure 2. It can clearly be seen that, for all the samples, the flow stresses present a typical dislocation slip dominated deformation feature, irrespective of the amount of pre-strain [25]. This suggests that almost no twinning or detwinning process takes place during the tension process [23,24]. Meanwhile, the mechanical properties of all three samples derived from the curves are listed in Table 1. Evidently, the yield stress of 0% pre-strained samples (ERD0CA) is the lowest, only about 178 MPa. As pre-strain increases, the yield stress also increases to 182 MPa and 186 MPa for ERD0.9CA and ERD1.3CA, respectively, which are only slightly higher than that of the samples without pre-strain. Interestingly, with increasing strain, the difference between the tension stresses of these three kinds of samples also increases. As indicated in Table 1, when the strain reaches 3%, the stress for ERD1.3CA is higher than that for ERD0CA, but not by much (~15 MPa). In addition, the elongations of these three kinds of sample are presented in Table 1. The elongation of ERD0CA is~21%, decreasing to~19% and~17% for ERD0.9A and ERD1.3A, respectively.

Mechanical Properties
Stress-strain curves under tension along ED for the ERD0.9CA, ERD1.3CA and ERD0CA are given in Figure 2. It can clearly be seen that, for all the samples, the flow stresses present a typical dislocation slip dominated deformation feature, irrespective of the amount of pre-strain [25]. This suggests that almost no twinning or detwinning process takes place during the tension process [23,24]. Meanwhile, the mechanical properties of all three samples derived from the curves are listed in Table 1. Evidently, the yield stress of 0% pre-strained samples (ERD0CA) is the lowest, only about 178 MPa. As pre-strain increases, the yield stress also increases to 182 MPa and 186 MPa for ERD0.9CA and ERD1.3CA, respectively, which are only slightly higher than that of the samples without pre-strain. Interestingly, with increasing strain, the difference between the tension stresses of these three kinds of samples also increases. As indicated in Table 1, when the strain reaches 3%, the stress for ERD1.3CA is higher than that for ERD0CA, but not by much (~15 MPa). In addition, the elongations of these three kinds of sample are presented in Table 1. The elongation of ERD0CA is ~21%, decreasing to ~19% and ~17% for ERD0.9A and ERD1.3A, respectively.   Table 1. Yield stress (0.2% proof stress) and stresses corresponding to different strains and elongation of the samples with and without pre-strain during tension along extrusion direction. Before tension, all these samples are annealed. The pre-strains for the three kinds of samples are 0 (ERD0CA), 0.9% (ERD.09CA) and 1.3% (ERD0CA), respectively.

Microstructure Features Caused by Pre-Strain
It is well known that 1012 extension twins are readily formed in magnesium alloys when the external compression load is perpendicular to the c-axes of the grains. According to the pole figures shown in Figure 1b, the (0001) poles of the initial material used in the present work are largely perpendicular to the ED and randomly distributed around the ED. Thus, when pre-compression along ERD takes place, twins may be formed theoretically in the grains, whose c-axes largely deviate from ERD. The EBSD maps shown in Figure 3a-c exhibits a typical microstructure feature of ERD1.3CA. Apparently, a number of 1012 twins, whose boundaries are colored in red, are indeed produced in such samples. In the current EBSD analysis, the tolerance within ±5 • of theoretical misorientation and axis (86 • < 1120 >) was acceptable. The other grain boundaries with misorientation angle values > 10 • are denoted in black. Figure 3b-d displays the EBSD maps using an  Figure 3a-c, respectively. Obviously, within some grains (such as grains 2 and 5), no twins occur. This is because the c-axes of these grains are nearly parallel to the ERD, as indicated by the inserted crystalline structure cells, which is unfavorable for the formation of twins [14,15]. Meanwhile, the corresponding (0001) pole figures of Figure 3b-d are also inserted. More interestingly and importantly, it can be seen that the (0001) pole figures of these pre-strained samples are strongly perpendicular to the ED and randomly distributed around the ED, almost identical to that of the initial material (see Figure 1b).  For a better understanding of the texture feature of these pre-strained samples, two grains containing twinned regions (grain 1 and grain 4) in Figure 3 are selected to perform texture examination, as presented in Figure 4. The IPF coloring EBSD maps of grain 1 and grain 2 are shown in Figure 4a,b, and the corresponding (0001) pole figures are exhibited in Figure 4c,d, respectively. In these images, the matrix regions and twinned areas are denoted as M i (i = 1, 3, 4) and T j (j = 1, 2, 3, 4, 5), respectively. When combining Figure 4a-c, the green circle representing the crystallographic orientation of M1 is located at the edge of the (0001) pole figure, indicating the c-axis of M1 is almost perpendicular to ED. Interestingly, after twinning, the crystallographic orientation of T1 (denoted as the blue star) still lies at the edge of the (0001) pole figure (see Figure 4c). Additionally, similar changes to crystallographic orientation can be found in M1 and T2, M3 and T3, and M4 and T4. These results unambiguously suggest that the textures before and after pre-compression along ERD are almost unchanged, which also agrees with the texture feature shown in Figure 3 and the observations in reference [26]. On the other hand, it should also be mentioned that, after pre-strain, some twinned regions, whose c-axes are not perpendicular to ED any longer, also exist in the pre-compression samples, just like T5 in Figure 4b (indicated by  Figure 4d). Such twinned areas influence the texture feature to some extent and may further affect the macroscopic performance of materials. However, the statistical results (not shown here) from about 800 grains reveal that such twinned areas are infrequent; only about 86 grains contain such twinned regions, resulting in the textures not changing much before or after pre-compression. This further indicates that the changes in mechanical properties caused by the texture associated with such twinned regions may be insignificant.

Microstructures Investigated by TEM
The bright field TEM images in Figure 5a exhibit the microstructure features of sample ERD1.3CA. The electron beam is nearly along the [1120] zone axis. In this case, the boundaries of 1012 twins are almost parallel to the electron beam, i.e., edge-on. In Figure 5a, one deformation twin can be found, which should be formed during the precompression process. The selected-area electron diffraction (SAED) pattern taken from the region containing the twin and matrix is inserted, indicating that the twin observed here belongs to the 1012 twinning mode. This is also consistent with the observations in Figure 3. Moreover, as can be seen in Figure 5a, both the twin and matrix in ERD1.3CA seem to be very clean, indicating that the density of dislocations should be low, which is ascribed to the annealing process. The HRTEM image shown in Figure 5b Figure 5a, both the twin and matrix in ERD1.3CA seem to be very clean, indicating that the density of dislocations should be low, which i ascribed to the annealing process. The HRTEM image shown in Figure 5b

The Influence of Pre-Strain on the Tension Property
The influence of pre-induced {10 12} twins on the mechanical properties of mag nesium alloys has been investigated extensively [14][15][16][17][24][25][26][27]. As shown in Figure 2 an Table 1, the yield stress (0.2% proof stress) of ERD1.3CA was the highest, about 186 MPa while it was only slightly higher (~8 MPa) than that of ERD0CA (~178 MPa). This seem to indicate that the boundaries of pre-induced {10 12} twins play an insignificant rol in the strengthening of magnesium alloy. However, on the other hand, it is worth notin that only about half of the grains could be twinned during pre-deformation in the cur rent work, since the {10 12} twins prefer to be formed within the grains with the c-axe away from the external load direction about 45~90°° (acute angle) [8]. If these TBs ca strengthen the grains, the yield stress of those grains with twins will be enhanced. Wit this in mind, when the grains without twins enter the stage of plastic strain, the grain with twins may be still at the stage of elastic deformation. Consequently, the yield stresse of pre-strained samples shown in Figure 2 may mainly be attributed to the plastic defor mation of those grains without twins.

The Influence of Pre-Strain on the Tension Property
The influence of pre-induced 1012 twins on the mechanical properties of magnesium alloys has been investigated extensively [14][15][16][17][24][25][26][27]. As shown in Figure 2 and Table 1, the yield stress (0.2% proof stress) of ERD1.3CA was the highest, about 186 MPa, while it was only slightly higher (~8 MPa) than that of ERD0CA (~178 MPa). This seems to indicate that the boundaries of pre-induced 1012 twins play an insignificant role in the strengthening of magnesium alloy. However, on the other hand, it is worth noting that only about half of the grains could be twinned during pre-deformation in the current work, since the 1012 twins prefer to be formed within the grains with the c-axes away from the external load direction about 45 • ∼ 90 • (acute angle) [8]. If these TBs can strengthen the grains, the yield stress of those grains with twins will be enhanced. With this in mind, when the grains without twins enter the stage of plastic strain, the grains with twins may be still at the stage of elastic deformation. Consequently, the yield stresses of pre-strained samples shown in Figure 2 may mainly be attributed to the plastic deformation of those grains without twins.
In order to understand the contribution of pre-induced 1012 TBs to the strength better, the stress values corresponding to larger strains were also extracted from the stressstrain curves, as shown in Table 1. It is conceivable that, with continued loading, those grains containing twins should also be involved in the deformation and accommodate the strain. Thus, the stress corresponding to a larger strain after yielding may reflect the influence of these pre-induced 1012 TBs themselves on the tension property. As can be seen in Figure 2 and Table 1, the distinction between the strength of samples with and without pre-induced TBs gradually increases with increasing strain, suggesting that the pre-induced 1012 TBs can affect the hardening behavior of magnesium alloy to some extent. Meanwhile, it is worth noting that, at the same strain, the stress of ERD1.3CA is always higher than that of ERD0.9CA and ERD0CA, indicating that the strengthening effect is also closely related to the level of pre-strain, which determines the amount of pre-induced 1012 TBs. When the strain is about 3%, the tension stresses of ERD1.3CA (~219 MPa) are obviously higher than those of ERD0CA (~204 MPa) and ERD0.9CA (~213 MPa). It is assumed that when the strain reaches about 3%, the plastic deformation has already taken place within all of the grains. Thus, it may be concluded that the 1012 TBs themselves can indeed enhance the strength of magnesium alloy, but limitedly, and the strengthening effect is related to the amount of pre-strain.

Strengthening Related to Twinning Boundaries
It is well known that the yield stress can be enhanced by introducing common highangle grain boundaries (HAGBs) into the metals, according to an empirical equation, i.e., Hall-Petch relationship [28] σ y = σ 0 +kd −1/2 (1) where σ y is the experimental yield stress of material, σ 0 is the friction stress when dislocation glides on the slip plane, d is the effective grain size and k is a constant related to the material itself. Inspired by this, 1012 TBs are introduced in magnesium alloys as new grain boundaries in some publications to reduce the effective grain size and the mean free path of dislocations, leading to the enhancement of yield strength [14][15][16][17]. Our present work shows that pre-induced 1012 TBs can indeed enhance the strength of magnesium alloys, especially their peak stress. It is assumed that every introduced TB can be considered as an "HAGB" in the present work; the average grain size of the twinned grains is estimated to be about ∼ 5 µm. In reference [8], Y.C. Xin et al. proposed that, in the case of tension along ED of AZ31 rod, σ 0 varied around 123 MPa and the value of Hall-Petch slope (k) was about 245 MPa·µm −1/2 . Thus, the corresponding stress which yields these grains can reach as high as ∼ 240 MPa, which is much higher than that caused by the pre-induced 1012 TBs (see Table 1). This seems to indicate that, compared to HAGBs, the blocking effect of 1012 TBs for the matrix dislocations is much weaker. Yu et al. [29] suggested that the weaker blocking effect of 1012 TBs mainly originated from the much higher misorientation of 1012 TBs (about 86 • ) than HAGBs. In addition to misorientation, the interfacial features of twins and grains are also remarkably different. In general, the atoms located at traditional HAGBs are usually chaotic [30], but the boundaries of 1012 twins consist of coherent boundaries and basal-prismatic interfaces, and the atoms on these boundaries are usually regular and orderly, as shown in Figure 5b. Such interfacial characteristics of 1012 twins were also observed in [22,31]. Presumably, the interfacial atom arrangements should also be responsible for the blocking effect. When the mobile matrix dislocations encounter general HAGBs, they will be strongly hindered and have to halt due to the distorted interfacial atom arrangement. In contrast, when the mobile matrix dislocations meet 1012 TBs, the process of absorption or transformation would take place, rather than strong blocking und that, when the matrix dislocations interacted with 1012 TBs, almost all of these matrix dislocations were absorbed or transmuted by TBs, rather than piled up in the front of TBs.
It should also be noted that, for the samples with pre-induced 1012 twins, if detwinning takes place during the following tension process, the corresponding strength would be reduced due to the low activation stress required by detwinning [13]. Based on their experimental observations, Li et al. [26] found that the yield stresses of the samples with and without pre-induced 1012 twins were almost equivalent and ascribed such a phenomenon to the detwinning process. In the current work, detwinning may play an insignificant role in reducing the strength caused by the pre-induced 1012 twins. This is because the c-axes of twinned areas are nearly perpendicular to ED in the pre-strained samples (see Figures 3 and 4), which is unfavorable for the activation of detwinning during the subsequent tension along ED [14,24]. As presented in Figure 2, no yield plateau, which is a typical feature of 1012 twinning or 1012 detwinning, can be found indicating that almost no detwinning process takes place during the tension along ED for ERD0.9CA and ERD1.3CA. Moreover, the microstructural features of ERD1.3CA-ED1.2T are also presented in Figure 6. Abundant 1012 twins can be identified here whose boundaries are colored in red. More importantly, almost no detwinning traces (as observed in reference [26]) can be found in the band contrast map shown in Figure 5a, further indicating that the detwinning process in ERD1.3CA-ED1.2T is not dramatic.
found indicating that almost no detwinning process takes place during the tension alon ED for ERD0.9CA and ERD1.3CA. Moreover, the microstructural features of ERD1.3CA ED1.2T are also presented in Figure 6. Abundant {10 12} twins can be identified her whose boundaries are colored in red. More importantly, almost no detwinning traces (a observed in reference [26]) can be found in the band contrast map shown in Figure 5a further indicating that the detwinning process in ERD1.3CA-ED1.2T is not dramatic.  Figure 5b is inserted at the right upper corner.

Conclusions
To conclude, the contribution of pre-induced {10 12} TBs to the tension propert of magnesium alloy has been investigated using a combination of EBSD technology an mechanical tests. A comparative study about the strength of samples with and withou pre-induced {10 12} twinning boundaries, but possessing an almost identical textur feature, was carried out. The experimental results show that the pre-induced {10 12}  Figure 5b is inserted at the right upper corner.

Conclusions
To conclude, the contribution of pre-induced 1012 TBs to the tension property of magnesium alloy has been investigated using a combination of EBSD technology and mechanical tests. A comparative study about the strength of samples with and without preinduced 1012 twinning boundaries, but possessing an almost identical texture feature, was carried out. The experimental results show that the pre-induced 1012 TBs can indeed improve the tension property of magnesium alloy, especially its peak stress, but limitedly. The weak strengthening effect of 1012 twinning boundaries is thought to be closely related to their interfacial structure characteristics. Compared to general high-angle grain boundaries, the atoms located on the twinning boundaries are more orderly. These findings are expected to provide a new understanding of the twinning behaviors of magnesium and may help to design and develop magnesium alloys with superior properties.