Effect of Loading Direction on Tensile-Compressive Mechanical Behaviors of Mg-5Zn-2Gd-0.2Zr Alloy with Heterogeneous Grains

Abstract

The tension-compression yield asymmetry caused by the strengthening of Mg-Zn-Gd-Zr alloy due to extrusion deformation is an important issue that must be addressed in its application. In this study, the effects of loading direction on the tensile and compressive mechanical behaviors of Mg-5Zn-2Gd-0.2Zr alloy were systematically investigated. As the loading angle (the angle between the loading direction and the extrusion direction) increases from 0° to 30°, 45°, 60° and 90°, the tensile yield strength decreases more significantly than the compressive yield strength. Consequently, the tension-compression yield asymmetry is gradually improved. Additionally, the ultimate compressive strength decreases more markedly than the ultimate tensile strength with the increment of the loading angle. In tensile tests conducted at 0°, 30° and 45°, two distinct stages of decreasing strain hardening rates are typically observed. For the 60° and 90° tensile tests, one unusual ascending stage of strain hardening rate is observed. For all compressive tests, three stages of strain hardening are consistently noted; however, the increment in strain hardening rate caused by {10–12} extension twinning decreases with the increasing loading angle. A model combining loading angle and Schmid factor distribution was established. The calculated results indicate that the dominant deformation modes during the yielding process also vary significantly with the loading conditions. This clarification highlights the differences in yield strength variations between tension and compression. Finally, an analysis of the plane trace and crack propagation direction near the fracture surface reveals the fracture mechanisms associated with tensile and compressive tests at different loading directions. This study promotes understanding of the mechanical behaviors of Mg-5Zn-2Gd-0.2Zr alloy under different loading directions, and helps to thoroughly elucidate the anisotropic effects of texture on the mechanical properties of magnesium alloys.

1. Introduction

In the past decades, magnesium (Mg) alloys have been widely used as metal structural materials due to their low density and abundant storage on our planet [1]. The development of the modern transportation industry creates higher demands of both high strength and high ductility of Mg alloys, while maintaining low production costs. For this reason, Mg alloys with new or heterogeneous structures have been developed in recent years to meet the high demands [2].

Generally speaking, heterogeneous structured Mg alloys are an emerging class of structure that consist of soft and hard domains which exhibit significant differences in mechanical or physical properties [3]. As heterogeneous structured Mg alloys exhibit significant improvement in strength while keeping good ductility when compared with those of homogenous structured Mg alloys, many studies have been conducted on how to prepare a heterogeneous structure in Mg alloys. These include extrusion [4,5], hard-plate rolling [6], equal channel angular pressing [7], surface mechanical attrition treatment [8], and friction stir processing [9]. Among these methods, extrusion is the most promising method for industrial application due to the fact that extrusion is cost-effective and can fabricate Mg alloys parts with large dimensions [10]. However, extruded heterogeneous structured Mg alloys often contain coarse deformed grains and fine dynamic recrystallized (DRXed) grains [11], which means that a strong extruded fiber texture will be formed in such structures. It is well known that fiber texture leads to mechanical property anisotropy along different directions in extruded bars [12], and this will be harmful for the formability and application of Mg alloys [13]. However, up to now, there has been relatively little research on mechanical anisotropy in extruded heterogeneous structured Mg alloys.

In comparison, a lot of research has examined the mechanical property anisotropy in extruded Mg alloys with a homogenous structure. Two representative works are selected as follows: S. Kleiner et al. investigated tensile mechanical properties along different directions in extruded AZ61 bars, and they discussed the variation of tensile strength by calculating the Schmid factor (SF) [14]. In recent years, Wu et al. studied the tension-compression yield asymmetry (TCYA) in AZ31 extruded bars. The SF calculation was also used to explain the variation of the TCYA in their study. Furthermore, compared to S. Kleiner et al.’s work, they linked the variation of activated deformation modes with the loading direction, which provided an intuitive explanation of strength anisotropy in the extruded AZ31 bar [15]. To sum up, previous works provide us with vital clues that SF calculation will also be helpful in explaining mechanical anisotropy of extruded heterogeneous structured Mg alloys. However, no study to date has taken the particular HCP structure of the Mg unit cell into consideration, which may lead to inaccuracy in the explanation of mechanical anisotropy. In addition, mechanical anisotropy not only contains yield strength, but ultimate strength, and fracture strain and strain hardening behavior are also included [16]; both of these should all be taken into consideration for a systematic study.

Therefore, in this paper, the Mg-5Zn-2Gd-0.2Zr alloy was selected due to its comparative mechanical properties and low cost [17]. The Mg-5Zn-2Gd-0.2Zr alloy was extruded into a bar to obtain a heterogeneous structure with coarse deformed grains and fine DRXed grains. The tensile and compressive mechanical behaviors, including strength, fracture strain and strain hardening rate were systematically studied, along different loading directions, and then a detailed discussion of the mechanical anisotropy of a heterogeneous structured experimental alloy are presented.

2. Experimental Procedures

2.1. Sample Preparations

The detailed preparation process of the casting experimental alloy has been documented in our previous work [17]. The casting experimental alloy was homogenized at 505 °C for 16 h, then machined to the billet of Φ100 × 350 mm. Before extrusion, the billet should be preheated for 2 h at 260 °C. In order to obtain a heterogeneous structure, a low extrusion temperature was used (260 °C), the extrusion ratio was set as 10:1, and the extrusion speed was 3 mm/s. The chemical composition of the extruded experimental alloy was determined to be Mg-4.58Zn-2.1Gd-0.18Zr by using an inductively coupled plasma analyzer (PerkinElmer Inc., U.S., Perklin-Elmer, Plasma 400).

2.2. Microstructure Characterizations and Mechanical Tests

Metallographic sample preparation included soft grinding on grinding discs ranging from 600, 1000, 1500, 2000 to 3000 grit SiC-paper. In order to avoid oxidation, ethanol instead of water was used for cooling during the whole grinding step. After grinding, 3, 1 and 0.5 µm oil-based diamond polishing was conducted on Struers MD-Nap cloths at a low rotation speed (50 rpm/min); polishing continued until scratches could not be recognized by using microscope. Electro-polishing in a 5:3 solution of ethanol and H₃PO₄ at 2 V was performed after diamond polishing, and the duration of electropolishing was set to be about 1 h. Electro-polished samples were rinsed with alcohol and then chemical color-etching was conducted by using a freshly prepared 1:1:7 solution of water, acetic acid and picral (4% picric acid in solution with ethanol). The etching time was set to be 5 to 10 s, and optical microscopy was obtained by using polarized light in a Leica DM2700 M light microscope.

The preparation process of samples for macro texture measurement was almost the same as that of the metallographic sample, the only difference being that etching was not used after polishing. A Bruker D8 Advance diffractometer equipped with a high-resolution area detector was used to characterize the macro texture in the extruded experimental alloy on different planes. Electron backscatter diffraction (EBSD) measurements were performed using a scanning electron microscope (SEM) equipped with a field emission gun (ZEISS Gemini 500) and an HKL-Nordlys II EBSD detector to characterize the microstructure. The preparation for the EBSD samples was also almost the same as for the metallographic sample; a commercial electrolyte (AC2) was used for electro-polishing. During EBSD tests, to ensure a perfect alignment towards the EBSD detector and increase the backscattering yield, samples were mounted onto pre-tilted (70° from the horizontal) brass holders by using conductive adhesives. An acceleration voltage of 20 kV, step size of 0.15 μm and working distance of 10 mm were used for all EBSD measurements. The raw EBSD data were analyzed by using commercial OIM Analysis 7.0 software. Before producing the inverse pole figure (IPF) map, clean-up procedures of all the raw EBSD data were carried by using grain dilation (clean-up parameters: grain tolerance angle = 5 and minimum grain size = 2), and EBSD data points with a confidence index (CI) lower than 0.1 were excluded from the analysis as they could be dubious (the CI quantifies the reliability of the indexed pattern).

Tensile and compressive tests were carried out in an INSTRON 3380 electronic universal material testing machine, and mechanical properties were measured at room temperature with a constant strain rate of 1 × 10⁻³ s⁻¹. A schematic diagram (Figure 1) shows tensile and compressive sampling positions in the extruded bar. It can be seen that the loading direction was set as 0° to 30°, 45°, 60° and 90° away from the extrusion direction (ED). The dog bone-like tensile sample was 5 mm in gauge length, 1.5 mm in width and 1 mm in thickness. The cylindrical compressive sample was 6 mm in diameter and 9 mm in height.

3. Results

3.1. Microstructure and Texture Under Different Loading Directions

Figure 2 shows optical micrographs selected from 5 different positions (0°, 30°, 45°, 60° and 90° plane, respectively), which have been depicted in Figure 1. It is clear that the microstructure showed different characteristics when observed from different planes. The microstructure obtained from the 0° plane shows a typical heterogeneous structure, which contains coarse grains and surrounding fine grains. With the increment of the angle, a heterogeneous structure is also observed in each plane, and the size of coarse grains decreases compared with that in the 0° samples. Such heterogeneous structures in extruded Mg alloys have already been reported in previous studies, in which a low extrusion temperature or ratio were often used [4,18]. As the extrusion ratio used in this study is 10:1, a heterogeneous structure containing un-DRXed coarse grains elongated along ED and fine DRXed grains are expected in the extruded bar.

Figure 3 shows the {0002}, {10$\overline{1}$0} and {11$\overline{2}$0} pole figures testing from different sampling planes. It is obvious that each pole varies significantly with the variation of the angle. Specifically, for the 0° sample, {10$\overline{1}$0} poles are located in the center of the polar diagram and have the highest intensity (Max = 8.3), showing a typical extruded fiber texture of extruded Mg alloys [19], and the c-axis of the Mg unit cell is distributed like a ring on the cross section of the extruded bar. As the angle increases, the circular distribution along the circumference of the {0002} polar graph gradually deflects towards the horizontal line of the polar graph, and the poles on the {0002} polar graph remain evenly distributed along the ring. The {10$\overline{1}$0} poles still exhibit the highest intensity, which remains almost constant as the angle increases (Max = 8.1, 8.3, 8.1, and 9.2), indicating good repeatability of the structure in the extruded experimental alloy.

3.2. Mechanical Properties of Tension and Compression Under Different Loading Directions

Figure 4 shows the engineering stress–strain curves of tensile and compressive specimens at room temperature under different loading directions. The results show that tensile (Figure 4a) and compressive (Figure 4b) mechanical properties and behavior are both affected by the loading direction. With the increase of the loading angle, tensile curves gradually change from the common parabola shape to an S shape, while compression curves under all loading angles show an S shape. Many previous studies have pointed out that the S-shaped curve in Mg alloys is closely related to the activation of twins [20,21,22], thus the shape change of mechanical curves may be closely related to the activation of twins for the present case, and the activation of twins under tensile conditions may be different from that under compression.

Figure 5 displays the tensile and compressive mechanical properties data under different loading directions. As can be seen from Figure 5a, the tensile yield strength (σ_yt) shows a trend of significant decrease with the increase of the angle, while the compressive yield strength (σ_yc) decreases at a relatively slower rate with the increasing loading angle. In this paper, values of σ_yc/σ_yt are used to measure the TCYA under different loading directions [23]. The calculated values of σ_yc/σ_yt at 0°, 30°, 45°, 60° and 90° are 0.74, 0.92, 1.01, 1.10 and 1.07, respectively. The calculated results show that the TCYA is most significant for the heterogeneous structure when loaded along the extrusion direction. With the increase of the angle, the TCYA of the alloy is greatly improved. When loaded along 45° or 90°, the TCYA almost disappears. When the loading angle is 60°, the compressive yield strength even exceeds the tensile yield strength.

The variation of ultimate tensile and compressive strength (σ_ut, σ_uc) are shown in Figure 5b. It can be seen that the σ_ut shows little change with the increase of the angle. The σ_uc decreases significantly with the increase of angle from 0° to 60°, but its value remains almost unchanged from 60° to 90°. Furthermore, it is also seen that σ_uc is higher than σ_ut under every loading angle. For the fracture strain, the tensile (ε_ft) and compressive (ε_fc) fracture strain show similar variation trends with the increase of the angle, in that they both increase at first and then decrease. ε_ft reaches its peak value at 45°, and ε_fc reaches its peak value at 60°, which is very close to the value at 45°.

Previous studies have concluded that strain hardening behavior during the uniaxial loading process was strongly associated with the tensile and compressive strength [24]. To study the variation of strength in this study, tensile and compressive strain hardening rates under different loading directions were calculated. The strain hardening rate (θ) -true strain curves are plotted in Figure 6, and different background colors represent different strain hardening stages. For tensile tests, 0°, 30°, and 45° specimens show only two stages of strain hardening rate. Specifically, the first stage of rapid decline in the stain hardening rate (gray background) is caused by the rapid elastic–plastic transition [25]. The next stage is a linear decrease of the stain hardening rate (corresponding to the blue background) and then a sharp decrease until plastic instability occurs. Once the loading angle increases to 60° and 90°, it can be seen that there is a stage where the strain hardening rate increases (marked by red background in the figure). This phenomenon is consistent with the change in shape of the tensile curves.

The compressive strain hardening rate-true strain curves show the following three stages. For all loading angles, the θ decreases rapidly at first (gray background), then it increases to a peak value (red background), and it decreases linearly then decreases sharply until failure (blue background). Closer observation further reveals that there are some differences for the ascent stage under each loading angle. The ascending stage of the 0° sample is the most significant, and the ascending range gradually decreases with the increase of the angle. Some researchers have pointed out that the increment of strain hardening rate in Mg alloy is closely related to the formation of twins [26,27]. Therefore, it can be inferred here that the difference in the ascending stage should be related to the activation of twins in the process of tensile and compressive tests under different loading angles.

3.3. Tensile and Compressive Fracture Under Different Loading Directions

SEM morphologies of tensile fracture surface under different loading directions are shown in Figure 7, Figure 7a–e are the macro fracture morphologies under low magnification, and Figure 7f–j are the local magnification morphologies under high magnification. It can be seen from these figures that fracture morphologies of tensile fracture surface under different loading directions are different from each other. For the 0° specimen (Figure 7a,f), a small number of smooth cleavage planes can be seen, which are surrounded by some dimples, indicating that the 0° sample mainly exhibits brittle fracture accompanied by few ductile fractures. With the increase of the loading angle, including the 30°, 45° and 60° samples, more and more cleavage planes are observed on the fracture surface, but the cleavage planes are not as smooth as those from the 0° sample. This indicates that these samples show the characteristics of shear fracture during the process of tensile failure [28]. For the 90° sample, smooth cleavage planes are observed again on the fracture surface, indicating that the tensile failure of the 90° sample is also dominated by shear fracture.

SEM morphologies of the compressive fracture surface under different loading directions are shown in Figure 8. It can be seen that the fracture morphologies of compressed samples under all loading directions are almost with the same as each other. Both samples are sheared along a direction of 45° from the loading axis, and a large area of cleavage planes and thin ridges is visible on the fracture surface. These features prove that the alloy exhibits typical shear fracture characteristics when compressive fracture occurs under each loading direction [29].

4. Discussion

The above results have provided detailed information on the mechanical behaviors of Mg-5Zn-2Gd-0.2Zr alloy with a heterogeneous structure, and it can be seen that mechanical behaviors vary significantly along different loading directions. However, there are still two issues that need to be explained. The first is the variation of mechanical strength, and the second is the variation of fracture behaviors. They will be discussed in detail in the following sections.

4.1. Revealing the Variation of the Strength Along Different Loading Directions

The Experimental alloy with a heterogeneous structure exhibits strong fiber texture, and SF is highly dependent on the orientation [30]. Thus, theoretical calculation of SF was carried out to reveal the variation trend of yield strength. Information about calculating SF distribution is provided as the Supplementary Materials in Supplementary S1 [12,15,30,31,32,33,34,35,36]. Tensile and compressive tests are discussed respectively. As shown in Figure 9a,b, for tensile tests in the 0° sample, the CRSS/SF of prismatic slip is the lowest among all deformation modes, and thus prismatic slip dominates the tensile yielding process. When the loading angle increases to 30°, it can be deduced that basal slip is activated during the tensile yielding process due to the increment of the SF of basal slip. As the CRSS of basal slip is much lower than that of prismatic slip, a sharp decrease of σ_yt will occur [36]. When the loading angle increases to 60°, it can be seen that the CRSS/SF further decreases, corresponding to the further reduction of tensile yield strength. The reason for this can be attributed to the fact that the SF of basal slip increases with the loading angle and thus, basal slip with low CRSS/SF gradually dominates the tensile yielding process. As a result, even though the 30° and 60° loading direction is symmetric about the 45° direction, σ_yt of the 30° sample is much higher than that of the 60° sample. When the loading angle increases from 60° to 90°, it can be seen that the SFs of each deformation mode gradually converge. As the CRSS of basal slip and {10$\overline{1}$2} twinning show similar values, it is reasonable to deduce that, apart from basal slip, {10$\overline{1}$2} twinning will also take part in the tensile yielding process. Therefore, tensile yield strength varies little with the loading angle increase from 60° to 90°, and the shape of tensile curves changes from the common parabola shape to the S shape due to the activation of {10$\overline{1}$2} twinning.

Ultimate tensile strength (σ_ut) is dependent on the tensile yield strength and strain hardening behavior. The difference value between σ_ut and σ_yt is often used to assess the strain hardening ability [37,38]. J.D. Valle investigated the influence of texture on the strain hardening behavior of Mg alloys, and confirmed that the activation of prismatic slip leads to the reduction of strain hardening ability [37]. That is to say, even though the 0° sample shows the highest yield strength, the activation of the prismatic slip results in the most significant reduction of the strain hardening rate in the second stage. With the increase of the loading angle, more and more basal slip will take part in the deformation process, leading to the slower reduction of the strain hardening rate. For the 60° and 90° samples, activation of {10$\overline{1}$2} twinning results in the increment of the strain hardening rate, which has been widely proved in previous studies [39,40,41]. To conclude, even though σ_yt decreases with the increment of the loading angle, the strain hardening ability is improved at the same time. Consequently, σ_ut varies little with the increment of the loading angle.

For compressive tests, as there is no polarity for dislocation slipping, the activation of slip systems is the same as those under tensile tests, and the only difference is the activation of {10$\overline{1}$2} twinning. CRSS/SF of {10$\overline{1}$2} twinning is the lowest among all deformation modes, and thus {10$\overline{1}$2} twinning dominates the compressive yielding process of the 0° sample. When the loading angle increases from 0° to 45°, more and more basal slip will occur in the compressive yielding process, leading to the reduction of σ_yc. As basal slip and {10$\overline{1}$2} twinning have similar values of CRSS, the reduction of σ_yc is inconspicuous. The 60° sample it shows same value of SF of {10$\overline{1}$2} twinning as that of the 30° sample, whereas the SF of basal slip is higher than that of the 30° sample. As a result, σ_yc further decreases at the loading angle of 60°. When the loading angle further increases to 90°, all deformation modes show an approached value of SF, and basal slip and {10$\overline{1}$2} twinning will dominate the compressive yielding process, leading to the further decrease of σ_yc.

In relation to σ_uc, the σ_yc varies less significantly than σ_yt. In this case, the twinning induced strain hardening rate determines the difference in value between σ_uc and σ_yc. Due to the dominant {10$\overline{1}$2} twinning, the 0° sample show the most significant effect of strain hardening. When the loading angle increases from 30° to 45°, basal slip will take part in the deformation process, leading to the reduction of strain hardening rate. As a result, σ_uc slightly decreases. The 60° sample shows the same strain hardening ability as that in the 30° sample, and σ_uc stays almost the same. Once the loading angle further increases to 90°, due to the slight decrease of strain hardening ability and σ_yc, a slight decrease of σ_uc is expected. Consequently, σ_uc varies significantly with the increment of the loading angle.

Based on the above discussion on the yield strength, the variation trend of TCYA can be revealed as follows. For the 0° sample, the strong fiber texture leads to different dominant deformation modes during the tensile and compressive yielding process. Thus, a strong TCYA is obtained. When the loading angle increases to 30°, more and more basal slip is activated during the tensile and compressive yielding processes, leading to the reduction of the TYCA. For the 45° sample, there is almost no difference between the activated deformation modes during the tensile and compressive yielding processes. When the loading angle increases to 60°, it can be deduced from the SF that {10$\overline{1}$2} twinning may contribute a higher portion to the compressive yielding than the tensile yielding. This is because CRSS of {10$\overline{1}$2} twinning is a bit higher than that of basal slip, leading to a reverse TCYA in the 60° sample (σ_yc is higher than σ_yt). At the loading angle of 90°, there is also no difference between the activated deformation modes during the tensile and compressive yielding processes. Consequently, the TCYA of the 90° sample is almost eliminated.

4.2. Tensile and Compressive Fracture Mechanisms Under Different Loading Directions

According to the SEM images of fracture surface (shown in Figure 7 and Figure 8), it is found that the fracture morphologies of tensile samples are different from each other under different loading directions, while the fracture morphologies of compressive samples are almost the same. To further analyze fracture mechanisms, EBSD is used to analyze the crack initiation, which was selected from the area near fracture surface. As the fracture area experienced large deformation, and to improve the credibility of EBSD analysis, points with CI value lower than 0.1 are removed during EBSD data processing.

For tensile fracture mechanism analysis, Figure 10 displays the crack analysis results at the fracture area of the 0° sample. Figure 10a is the macroscopic SEM image near the fracture surface. An evident crack can be seen in the dotted red line box, which is further amplified in Figure 10b, showing that the crack propagation direction is parallel to the fracture surface. Figure 10c is the IPF map of the corresponding region, referring to the loading direction. It can also be found that the propagation direction of cracks is parallel to many thin lamellas, and the CI values within these thin lamellas are much lower than in the matrix, indicating that large residual strain lies in these areas. Moreover, these thin lamellas are found to be formed in deformed grains with <10$\overline{1}$0> orientation. One thin lamella marked with white dotted box is selected for further analysis, and its magnified image is shown in Figure 10d. It can be seen that the orientation relationship between the thin lamella and the <10$\overline{1}$0> oriented grains is 36° <11$\overline{2}$0>, which agrees well with the orientation relationship of the {10$\overline{1}$1}–{10$\overline{1}$2} secondary twinning. Therefore, these thin lamellas are identified as {10$\overline{1}$1}–{10$\overline{1}$2} secondary twin lamellas. According to the study by M.R. Barnett et al., due to large stresses at the final stage of the tensile test and high SF (seen in Figure 9a), {10$\overline{1}$1} contraction twinning will be activated within coarse grains with <10$\overline{1}$0> orientation [42]. The contraction twinning part is turned into the orientation of 56.3° <11$\overline{2}$0> with the matrix [42]. The orientation of contraction twinning grains is conducive to the further {10$\overline{1}$2} twinning under the tensile stress. Therefore, many {10$\overline{1}$1}–{10$\overline{1}$2} secondary twinning lamellas are found here [43]. A. Chakkedath et al. reported that {10$\overline{1}$1}–{10$\overline{1}$2} secondary twinning is prone to local deformation by using atomic simulation, thus twin boundaries can not move freely [44]. As a result, thin secondary twin lamellas will be formed. Plane traces of the {11$\overline{2}$0} matrix are depicted in Figure 10d, and it is clear that the crack propagation direction is parallel to the {11–20} plane of the matrix. Based on the above discussion, the tensile fracture mechanism of the 0° sample is such that secondary twins are formed within coarse deformed grains at the late stage of the tensile test; such twin lamellas cannot grow freely and this leads to stress concentration within these thin lamellas. Once the shear fracture stress of the Mg matrix is achieved, a crack will be formed and this propagates along the {11$\overline{2}$0} plane. That is the reason why many smooth cleavage planes are observed in the fracture surface (seen in Figure 7a).

Figure 11a shows a macroscopic SEM image near the fracture surface of the 30° sample. Fewer cracks are observed than those of the 0° sample, which is consistent with the significantly improved tensile elongation. Figure 12b shows the enlarged morphology of the two small cracks marked by the red dotted lines in Figure 11a. The two small cracks are labelled as Crack A and Crack B, respectively. Figure 11c is an IPF map of the corresponding region, and Figure 11d,e are enlarged maps of Crack A and Crack B, respectively. Black areas are formed due to points with low CI value. As shown in Figure 11d,e, thick red lines represent the propagation direction of cracks, and plane traces are depicted by thin lines. By comparing plane traces of grains near Crack A and Crack B (shown as thin lines in the figure), it is found that the direction of crack growth is well matched with the trace of the {10$\overline{1}$0} prismatic plane. As the prismatic slip shows a comparatively low CRSS and high SF, it can be expected that prismatic slip will be activated during the last stage, when tensile along the 30°direction. Therefore, it is deduced that the fracture occurs along the {10$\overline{1}$0} prismatic plane due to the shear fracture caused by prismatic dislocation slip in the final stage of the tensile test.

Figure 12a shows a macroscopic SEM image near the fracture surface of the 45° sample. A small crack marked with dashed line can be found in the figure, and its propagation direction is also roughly parallel to the macroscopic fracture surface. Figure 12b shows an enlarged image of the crack selected from the dotted box in Figure 12a. Figure 12c shows an IPF map of the corresponding region, and Figure 12d is the magnified image of the crack area. As shown by the red thick line and thin trace lines in Figure 12d, the propagation direction of the crack matches well with the trace of the {10$\overline{1}$0} prismatic plane. Combined with the fracture surface observation in Figure 7c,h, the 45° sample also presents the characteristic of a shear fracture. Therefore, it can be inferred that the shear fracture is also caused by the prismatic plane slip in the final stage of the tensile test, leading to the propagation of cracks along the {10$\overline{1}$0} prismatic plane of the grain.

Figure 13a shows a macroscopic SEM image near the fracture surface of the 60° sample. Small cracks marked by a dotted box are also shown in the figure, and the propagation direction is also roughly parallel to the fracture surface. Figure 13b shows the enlarged morphology of two small cracks marked by a dotted box in Figure 13a, and these two small cracks are labelled as Crack A and Crack B. Figure 13c is the IPF map of the corresponding region, and Figure 13d,e are enlarged images of Crack A and Crack B, respectively. These two cracks penetrate through several surrounding grains. According to the analysis of crack and trace direction, the propagation direction of Crack A is consistent with the trace direction of the {0002} basal plane, while the direction of Crack B is well matched with the trace direction of the prismatic plane {10$\overline{1}$0}. Combined with the fracture analysis in Figure 7d,i, the tensile fracture of the 60° sample also presents the characteristic of a shear fracture. As the basal slip and prismatic slip show comparatively low CRSS and high SF, it can be expected that basal slip and prismatic will be activated during the last stage when tensile along the 60°direction. Shear fracture occurs when a crack propagates along the {0002} basal plane or {10$\overline{1}$0} prismatic plane of the grain, due to basal or prismatic slip.

Figure 14a shows a macroscopic SEM image near the fracture surface of the 90° sample. There are many secondary cracks on the surface of the specimen, which is consistent with the low elongation of the 90° sample. Figure 14b shows the enlarged image of two small cracks in the dotted box in Figure 14a. The two small cracks are labelled as Crack A and Crack B. Figure 14c is the IPF map of the crack area, and Figure 14d,e are the enlarged maps of Crack A and Crack B, respectively. These two cracks also penetrate through several surrounding grains. In Figure 14d, multiple deformation mechanisms were compared, but no trace of the slip plane of any deformation mechanism was found to be parallel to the propagation direction of the crack. Interestingly, multiple {10$\overline{1}$2} twinning variants are found to intersect with each other near the crack (red lines marking 86.3°<11$\overline{2}$0> grain boundaries). Therefore, it is speculated that the generation of cracks here is related to the stress concentration caused by the intersection of twins [45]. For Crack B, the propagation direction is well matched with the trace direction of the {10$\overline{1}$0} prismatic plane. Combined with the fracture analysis in Figure 7e,j, the tensile fracture of the 90° sample presents the characteristics of a typical brittle fracture, indicating that the sample is less affected by the shear fracture caused by dislocation slip. The brittle fracture characteristic and formation of smooth cleavage planes are mainly due to the stress concentration caused by intersection of {10$\overline{1}$2} twin variants.

The fracture morphology of compressed samples under different loading directions is similar; for simplicity, the 0° sample is selected for analysis. Figure 15a shows a macroscopic SEM image near the fracture surface of the 0° sample, and Figure 15b shows the enlarged image of a crack (red dashed box) in Figure 15a. The propagation direction of the crack is also roughly parallel to the macroscopic direction of the fracture, and the crack also penetrates several grains. Figure 15c is the IPF map of the crack region. It has been found in our previous study that {10$\overline{1}$2} twinning is the dominated deformation mode during compressive test of the extruded Mg alloy with fiber texture [25]. In this case, the boundaries of {10$\overline{1}$2} twins (86.3°<11$\overline{2}$0>) are marked with purple lines. There are almost no {10$\overline{1}$2} twins observed near the fracture surface, which is caused by the fact that the twinning completely consumes the matrix at the late stage of compression deformation. The color of the IPF map is all red, which is due to the reorientation caused by {10$\overline{1}$2} twinning [25]. Figure 15d is the enlarged map of the crack area, and the propagation direction of the crack matches well with the trace of the basal plane. Combined with the fracture analysis in Figure 8a,f, the compression fracture of the 0° sample presents obvious shear fracture characteristics. Basal slip is reported to be activated in the extension twinning parts of the parent grains with <10$\overline{1}$0> direction. Considering basal slip shows the lowest CRSS, it is deduced that basal slip is responsible for the compressive fracture. Therefore, the shear fracture occurs when the crack spreads along the {0002} plane of the grain due to the basal slip in the late stage of compression.

5. Conclusions

This study systematically investigates the effect of loading direction on the tension-compression mechanical behaviors of extruded Mg-5Zn-2Gd-0.2Zr alloy with heterogeneous grains. The study found that, under stretching and compression conditions, the different types of slip systems and twinning initiation lead to significant tension-compression yield asymmetry in the experimental alloys. The tensile properties of the alloy are determined by the deformation mechanisms of twinning, basal slip, or prismatic slip, while the compressive properties are determined by the basal slip mechanism. Finally, the following conclusions were drawn:

(1) With the increase in loading angle from 0° to 90°, the tensile yield strength exhibits a more obvious decrease compared to compressive yield strength. This trend indicates a gradual reduction in tension-compression yield asymmetry. There is more obvious decreasing of the ultimate compressive strength than the ultimate compressive strength. The tensile and compressive fracture strain show similar variation trends with the increase of the angle, in that they both increase at first and then decrease.

(2) For variation of strain hardening rate during tension, two declining stages are commonly observed in 0°, 30° and 45° samples, while an unusual ascending stage appears at 60° and 90°. For all compressive samples, three commonly observed stages of strain hardening rate are found, although the increment of strain hardening rate decreases with the increasing loading angle.

(3) A model combining loading angle and SF distribution is established. Based on the calculated results, it is found that the dominant deformation modes during the yielding process also vary significantly with the loading condition. The sharp decrease of tensile yield strength is related to the large disparity of the CRSS of dominant deformation modes during tensile yielding. Accordingly, the minor change of compressive yield strength is caused by the similar CRSS of dominant deformation modes during compressive yielding.

(4) According to the plane trace and crack propagate direction alignment near the fracture surface, fracture mechanisms during tensile and compressive tests along different loading directions are observed. The tensile fracture mechanism varies from shear fracture to brittle fracture when the loading angle increases from 0° to 90°, and the compressive fracture under each loading direction is dominated by shear fracture along a 45° angle away from the loading axis.