Effect of high-pressure torsion on microstructure, mechanical properties and corrosion resistance of cast pure Mg

The microstructures of the top surfaces in the cast and HPT-processed Mg samples after one and five turns are shown in the EBSD overlapped inverse pole figure (IPF) and grain boundary maps in Fig. 2 where the colors denote different grain orientations as depicted in the upper unit triangle. It can be seen that the cast Mg in Fig. 2a has a very coarse grain structure of a few millimeters with some twins whereas the grain sizes in the HPT-processed samples were refined significantly to ~ 1.6–4.0 µm in Fig. 2b–g for the centers, halfway positions and edges of the disks, respectively. A comprehensive summary is given in Table 1 of the grain sizes at the top surfaces for these three positions for both a combination of low-angle grain boundaries (LAGBs having misorientations below 15°) and high-angle grain boundaries (HAGBs with misorientations at and above 15°) and for the HAGBs only.

Figure 2b–g relates to the surface microstructures of the HPT-processed disks at these different radial locations within the circular samples and inspection suggests that they exhibit typical bimodal grain size distributions. The microstructures of both samples at the center after one and five turns in Fig. 2b, e are composed of a relatively large fraction of coarser grains mixed together with a small fraction of fine grains. By contrast, the fractions and sizes of the coarser grains are reduced at greater distances from the center. These results demonstrate, therefore, that the grain structure tends to become more uniform from the center to the edge along the radial direction and this is reasonable considering the increase in strain with increasing distance from the center.

According to Fig. 2, the HPT processing refined the grain structure and resulted in much refined equiaxed grains. It can be seen that, in addition to grains of several micrometers in size, there are much smaller grains of a few micrometers. This is most evident after one turn and at the center and therefore after the smallest strain (Fig. 2b) where very small grains are located in groups which are evidently within prior grains and they are not formed as necklace-like structures along the original grain boundaries of the large grains. None of these grain groups are red so that they are without the basal texture component. All very small grains have the same orientation, so that the grain boundaries between them are LAGBs and they correspond to subgrain boundaries. It is concluded that these grains are not recrystallized but they are dynamically recovered. Because the groups are absent after higher strains (from Fig. 2c onwards), although smaller and larger grains remain evident in the images, it seems that continuous dynamic recrystallization takes place in these grain areas at larger strains where the large grains are a consequence of discontinuous recrystallization.

A bimodal grain structure was found that was similar to that reported after HPT of Mg in earlier reports [15, 20]. For example, there was a bimodal microstructure including large recrystallized grains, free of dislocations, in the steady-state condition with an average grain size of about 1 μm [15] and there is another report of equiaxed grains with a size of about 1–5 μm [20]. Because the latter grain size increased with increased applied pressure, it was concluded that the grains are formed by both discontinuous static and dynamic recrystallization mechanisms. Thus, after 5 turns and 6 GPa pressure two different types of regions were reported, with “deformed regions” having ultrafine grains and a high density of dislocations and “recrystallized regions” with substantially larger grains (1–5 μm) and almost free of dislocations [21]. Thus, the bimodal grain structure observed in the present investigation has similarities to several earlier reports.

The sample processed by five turns also demonstrated a similar variation in the grain structure as in the one-turn sample. Furthermore, the grain size showed no significant change with increasing numbers of turns as documented in Table 1 except only at the disk center where there is no torsional straining and consequently there was only a high compressive load.

In the EBSD images in Fig. 2, it is readily apparent that most of the grains are red in color. This means that the {0001} planes are parallel to the disk surface and this preferred orientation is essentially maintained over the entire surface. Figure 3 displays the pole figures of the HPT-processed Mg at different locations after (a–c) 1 and (d–f) 5 turns. Thus, the pole figures after one turn at the center of the sample demonstrate a {0001} <10\( \bar{1} \)0> basal texture with an increasing intensity with increasing strain toward the edge of the sample. The ODF plots in Fig. 4 also confirm the texture evolution on the surface through the radius of the sample and by increasing to five turns of HPT a homogenous {0001} basal texture exists over the whole sample surface as is evident in Fig. 3. Since the as-cast Mg had coarse grains, and in the cross sections only two or three grains were visible, it was not possible to determine the texture.

In order to check the homogeneity of the grain structure through the cross sections of the HPT-processed disks, the microstructure over the whole thickness after one turn is presented in Fig. 5 in three radial locations. The orientation of the grains is viewed in two directions, the direction perpendicular to the disk surface (Fig. 5a–c) as well as in a direction parallel to the surface (Fig. 5d–f). The first images show red grains so that there is a strong basal texture through the disk parallel to the surface. The final three images (Fig. 5d–f) show bluish and greenish colors so that most of the grains have orientations close to (01\( \bar{1} \)0) and (\( \bar{1} \)2\( \bar{1} \)0) when looking toward the cross section of one-half of the disk. The microstructure through the thickness is not entirely homogeneous after one turn in the center region and there are a few flattened (about 100 μm long) grains containing LAGBs which indicate the occurrence of recovery in this region (see Fig. 5a, the encircled larger grains). Very small recovered grains of a few micrometers in size were also observed at the surface layer (in Fig. 2b) but due to a much lower resolution in Fig. 5a these grains cannot be resolved. It appears that dynamic recrystallization can occur in most grains already during one turn but also there are areas of dynamic recovery. Inhomogeneities through the cross sections of HPT-processed disks were also reported in some other investigations [22‐24]. It is noted that in the halfway and edge regions in Fig. 5b, c the grains appear to be more uniform in size and recrystallized, which is similar to the conclusions from Fig. 2c, d.

Figure 6 presents pictorial displays of the distributions of the area fractions of different grain sizes at the center and edge for both (a and b) one and (c and d) five turns. After one turn at the center in Fig. 6a, there is a distinct difference between the top and bottom layer, with the former having a higher fraction of fine (< 10 µm) grains and the latter having more numerous larger (> 30 µm) grains. This difference is not so pronounced at the edge in Fig. 6b and after five turns the grain sizes became more homogeneous in Fig. 6c–d and the grain size distributions through the thickness after 5 turns are similar in the top and bottom of the sample.

The microhardness of the samples was measured in order to estimate the changes in the microstructure and strength attributed to the HPT processing. Figure 7 shows the variation in Vickers microhardness as a function of distance from the center of the disk samples in the as-cast material and after 1 and 5 turns. The initial hardness of the as-cast Mg was relatively high, 41 Hv_0.01. The results show that the hardness increases clearly by one turn of HPT processing to about 60 Hv_0.01. However, increasing the turns to 5 slightly decreases the hardness. Thus, the hardness variation is irregular and any sign of a peak or any trend by distance from the centers of the disks is not generally apparent from these data. These results are in agreement with other observations [15, 25]. For example, there is a report that even after less than one turn there was an increase in hardness with respect to the distance from the center but at higher deformations the opposite occurred due to dynamic softening.

In practice, the hardness is affected by three separate factors: the finer grain size, the texture of the HPT samples and the higher dislocation density which is an inherent feature of the SPD-processed microstructure. In this study, the microhardness of HPT samples did not change by increasing the distance from the center although the grain size was slightly reduced. The basal texture is quite similar over the total surface as shown in Fig. 2, 3 and 4. Therefore, the dislocation density and thus the extent of recovery must vary with strain and have the main impact on the microhardness values. After one turn at the center, there are numerous subgrains and it is concluded that there is a higher dislocation density in an incompletely recrystallized structure which accounts for the higher hardness in this stage.

The grain boundary misorientation distributions displayed in Fig. 8 also confirm the existence of higher numbers of LAGBs, and therefore of recovered grains, after 1 turn compared with 5 turns, where it was shown earlier that the two main peaks, appearing at about 27° and 88°, correspond to the existence of doubled and tensile twins [26]. The fraction of LAGBs varies along the radial directions due to the different strain concentrations and in practice the fraction is higher at the center than in the edge region. Thus, after 5 turns a fully recrystallized structure was achieved with a lower hardness instead of a partly recovered structure.

It has been noted that a softening behavior tends to occur in pure metals when the homologous temperature is larger than ~ 0.3 [27] and in the present investigation the homologous temperature was ~ 0.32. It was observed by transmission electron microscopy that a replacement of the “deformed regions” in HPT Mg by dislocation-free recrystallized grains led to a softening [21] and it was suggested that, due to the faster defect recovery in Mg compared to metals such as Al, Cu and Ag, a saturation in hardening may occur in Mg at lower strains than in those other metals [25].

It was also noted earlier that additional HPT processing after 1/8 turn produced no significant increase in the microhardness [25] and this hardening at very small strains is probably due to the hexagonal close-packed (hcp) structure which provides basal slip and only two independent slip systems which are easily activated at room temperature. In order to fulfill the von Mises criterion, twinning and/or non-basal slip is also required at the onset of the HPT deformation [28].

Processing by SPD procedures is used conventionally to improve the strength properties of bulk solids. Accordingly, tensile testing was conducted both before and after HPT processing and representative engineering stress–strain curves for the cast and HPT-processed miniature samples are plotted in Fig. 9. These curves show that the yield and ultimate tensile strengths are markedly increased by HPT processing as listed in Table 2, but conversely, and as anticipated from the paradox of strength and ductility [29], the overall elongations are reduced through HPT processing. The concave tensile behavior of the cast Mg was explained previously [26]. Apart from the grain refinement and the dislocation density, texture strengthening also contributes to strength of Mg. In practice, texture strengthening is generally caused by deformation along the forming direction which limits basal slip and is beneficial for the activation of non-basal slip and thereby results in an increasing yield strength. It should also be considered that the low stacking fault energy on the basal plane (36 mJ m⁻²) may reduce the dislocation slip and cause strain hardening [30].

Corrosion resistance is one of the most important properties for biomedical applications since corrosion products may give some cytotoxicity effects and low corrosion resistance can cause a significant reduction in mechanical strength. Therefore, in this research the corrosion resistance of the samples was examined to determine whether HPT retards or facilitates the corrosion of Mg.

Figure 10 shows the Nyquist impedance plots for the cast and HPT-processed Mg when using NaCl solution. It can be seen that the plots for all samples include one capacitive loop and one inductive loop at low frequency. The capacitive loop has been assigned to charge transfer and film effects and the low-frequency inductive loop is related to the relaxation of adsorbed species such as Mg⁺ and non-stationary conditions [31]. The same shapes of the Nyquist plots, as well as the same diameters of the capacitive loops for all samples, suggest a similar corrosion mechanism and equal corrosion rates. Two earlier studies determined two time constants for Mg and its alloys immersed in NaCl [31, 32] where a reduction in the corrosion resistance of pure Mg was observed after processing by ECAP. Figure 11 shows the equivalent circuit that was used to fit the impedance data by ZSimpwin software and the results are listed in Table 3 where R_s is the electrolyte resistance between the working and reference electrode, C_dl is the electrochemical double-layer capacitance at the substrate/electrolyte interface and R_CT is the charge transfer resistance. For a best fit with the experimental data, the inductive part was omitted from the equivalent circuit and the fitting was carried out between 1 and 100 kHz.