Due to limitations in energy and environmental resources, lightweight has attracted increasing attention in industry [1,2]. As the lightest structural materials, Mg alloys are of considerable interest in the fields of transportation vehicles, electronic apparatuses and aerospace, due to their advantages of low density and high specific strength and stiffness [3]. Among various Mg alloys, Mg-Zn-Zr alloys exhibit an excellent combination of strength and ductility. However, like the other Mg alloys, the hexagonal close-packed (HCP) crystal structure and limited slip systems of Mg-Zn-Zr alloys cause poor formability at room temperature which restricts their commercial applications. Therefore, deformation processing is normally conducted at elevated temperatures to yield Mg alloy parts with superior mechanical properties.

Hot extrusion, the dominant plastic forming process of Mg alloys, can be used to produce solid and hollow profiles with improved microstructure and mechanical properties. Hollow profiles can greatly reduce the weight of the formed products [4,5], showing a better prospective. In industry, hollow profiles are usually produced by means of porthole die extrusion, during which a preheated billet is first forced into portholes and cleaved into several fresh metal streams by port bridges. Then, as the extrusion process continues the separated streams are solid bonded inside a welding chamber. Finally, the profiles are extruded through a die bearing, and several longitudinal weld seams are inevitably formed along the whole length of the profile. Weld seam is usually the weakest position, where cracks are likely to initiate under external stress [6]. Thus, welding quality of Mg profiles extruded by porthole die was investigated based on numerical simulations, experiments and physical simulations, and found that it was greatly affected by the billet temperature, extrusion speed and initial billet microstructure [4,5,7,8].

The formation of a peripheral coarse grain (PCG) structure is a common occurrence during the extrusion of nonferrous metals, such as Mg and Al alloys. PCG structures are also an important issue that should be paid significant attention, since it contributes adversely to the surface quality, strength, fracture toughness, fatigue resistance, stress corrosion susceptibility and machinability of materials. In recent years, some mechanisms were proposed for the formation of the PCG structure in extruded Al profiles and can be mainly divided into two theories, viz., dynamic abnormal grain growth (DAGG) and static abnormal grain growth (SAGG). The former proposes that PCG structure should form via the abnormal grain growth of the dynamic recrystallized (DRXed) grains, which occurred during plastic deformation, while the latter attributes the presence of PCG structure to the abnormal grain growth of static recrystallized (SRXed) grains formed after plastic deformation. Williamson and Delplanque [9] studied DAGG using the energy minimization recrystallization model and found that DAGG only occurred when the strain rate was very low. Hence, the DAGG theory are inadaptable to reveal the occurrence of the PCG structure on the material surface, which experienced a high strain rate during extrusion process. Based on the traditional SAGG theory, extruded profiles cannot be cooled down quickly from the extrusion temperature to a sufficiently low temperature, resulting in the enablement of abnormal grain growth and the formation the PCG structure. The formation process of PCG structure during the extrusion of Al alloys can be detailedly explained as follows. Equiaxed grains of homogenized billets were first elongated under the effects of shearing and friction stress. Then, a fibrous or banded structure, which was the surface microstructure of the profile when it was initially extruded through the die bearing, was formed by grain engulfing and growth. Meanwhile, the SRX process started and consumed the stored deformation energy. Finally, some grains exhibited abnormal grain growth by consuming other smaller grains. Geertruyden et al. [[10], [11], [12]] performed numerous studies to verify and improve the SAGG theory. They compared the surface microstructures between normal extruded profiles and extruded profiles interrupted with immediate quenching, showing that the PCG structure was only present in the profile extruded by the normal process, while it was absent in the profile obtained with immediate quenching. Additionally, PCG structure formed short-range order when annealing treatments were employed on water quenched profiles. Those abovementioned results again proved that the occurrence of PCG structure is a static process. However, the deformed microstructure of the extruded profiles at the die exit contained a fine, equiaxed grain morphology resulting from DRX. Therefore, a new SAGG mechanism was proposed, in which the as-deformed microstructure with a nearly complete DRXed structure provided a precursor for abnormal grain growth. Based on these two abovementioned theories, the complete recrystallization microstructure, no matter if it is formed by SRX or DRX, is the essential condition for the formation of the PCG structure in extruded Al profiles.

Several factors were indicated to affect PCG structure, such as the temperature, especially the exit temperature, extrusion speed, extrusion ratio, alloying elements and initial billet microstructure. Ko et al. [13] attempted to decrease the exit temperature of a 6061 Al alloy by die cooling with N₂ gas, which reduced the abnormal grain growth and enhanced the mechanical properties of the extruded profiles. Geertruyden et al. [10] carried out extrusion experiments under different extrusion speeds and extrusion ratios with various contents of alloying elements and found that increasing the extrusion speed/ratio or reducing the number of recrystallization-inhibiting elements could aggravate the abnormal grain growth. Hence, the control of PCG structure depends on the balance between the driving force for recrystallization from the stored deformation energy and the retarding force for grain growth from the pinning effects of second phase particles.

Mg alloys are low stacking fault energy metals, showing different recrystallization mechanisms from Al alloys. Moreover, the plastic deformation, material flow and microstructure evolution of porthole die extrusion processes are much more complicated than those of conventional flat die extrusion processes. Therefore, it is an important issue to clarify the formation and evolution of surface coarse grains during the porthole die extrusion of Mg alloys. In this study, both extrusion experiments and numerical modeling were carried out. Grain structure, texture and precipitates at different positions of the extruded profile were characterized, and the evolution mechanism of the surface coarse grains during extrusion and air cooling were summarized. Moreover, the effects of the surface coarse grains on the hardness, tensile properties and wear performance were discussed.

Billets of as-cast Mg alloy ZK60 (Mg-5.65Zn-0.66 Zr, in wt%) with diameter of 120 mm were used as raw material in the present study. After a homogenization treatment at 410 °C for 16 h and followed by air cooling to room temperature, the billets were used for porthole die extrusion on a 12 MN extrusion line as schematically depicted in Fig. 1. The setup mainly consists of an upper die, a lower die, a bridge and a container, which are all machined using H13 tool steel. Prior to extrusion, the container, dies and billets were pre-heated up to and held at 380 °C for 20 min. A constant ram speed of 0.5 mm/s was operated during the entire extrusion process. The extruded profiles were naturally cooled down to room temperature in air. A plate profile with a cross-sectional thickness of 4 mm and cross-sectional width of 70 mm was achieved, and the extrusion ratio was calculated to be 40.37. Extrusion direction, transverse direction and normal direction of the extruded profiles are abbreviated as ED, TD and ND, respectively.

Sampling position for microstructure observation, tensile test, hardness test and wear test is indicated in Fig. 1. As is seen, 5 plates with lengths of 200 mm were cut at positions of 0 m, 1 m, 2 m, 3 m, and 3.5 m away from die exit, which are named as P1, P2, P3, P4 and P5, respectively. The microstructure was characterized through scanning electron microscopy (SEM), electron back-scattered diffraction (EBSD) and transmission electron microscopy (TEM). For SEM observation, the samples were polished and then etched using a mixture of 4.2 g picric acid, 10 ml acetic acid, 10 ml distilled water and 70 ml alcohol. The EBSD samples were electropolished at 30 V for 40 s in a solution containing 800 ml ethanol, 100 ml propanol, 18.5 ml distilled water, 75 g citric acid and 10 g hydroxyquinoline (ACII electrolyte) with the cooling of liquid nitrogen, and the EBSD data were analyzed by Channel 5 software. For TEM observation, TEM foils with diameter of 3 mm were mechanically ground to about 50 μm, and then were ion milled using a Gatan precision ion polishing system with an accelerating voltage of 4.0 keV at an operating angle of 8° for 30 min to obtain an initial perforation. After that, such foils were further thinned using an accelerating voltage of 3.5 keV and an operating angle of 6° until a hole could be observed. The temperature was maintained at -145 °C during the entire process. The tensile test was carried out at ambient temperature with a speed of 0.25 mm/min, and the fracture surface was observed by SEM. The microhardness was tested by applying a load of 300 g with a dwelling of 10 s. Ball-on-disc type tribometer was adopted for the dry sliding wear test at ambient temperature, in which a steel ball with a diameter of 5 mm was used as the counter-face and it slid 72 m on the polished sample with a speed of 40 mm/s under the normal load of 40 N. Prior to and after the sliding tests, the samples were cleaned by acetone and dried to obtain the weight loss. Moreover, the worn surfaces were also observed using SEM.

Mechanical variables such as strain, strain rate and temperature significantly affect the evolution of microstructure. Hence, to better explain the experimental results, a numerical model based on HyperXtrude was employed to develop a finite element (FE) simulation of the extrusion experiment. The numerical model is composed of profile, bearing, porthole, welding chamber and billet as shown in Fig. 2. All the process parameters used in simulation were identical to those of the experiment, and a transient state analysis model was adopted. The boundary conditions for the material flow and heat transfer equations were treated as time-dependent [14]. Moreover, it was reported that the severe cohesive action existed between the preheated billet and container or die cavity, which resulted in a sticking friction condition [15]. However, the friction between the billet and die bearing turned to slipping condition, and the slipping friction became the dominant one [16]. Hence, the friction conditions on the interface of billet/container and billet/die were set as sticking friction, while the Coulomb friction model with a coefficient of 0.3 was set at the die bearing. In addition, the constitutive equation of a ZK60 alloy proposed by Mirzadeh [17] was used in the numerical model.

As is known, materials in welding and matrix zones experienced significantly different histories of temperature, strain, and strain rate, resulting in the inhomogeneous microstructure between welding and matrix zones [18]. To obtain an overall understanding of the microstructure features of the extruded Mg profile, the cross-section of P1 was selected for the EBSD analysis, and the results are shown in Fig. 3. The thick black lines indicate high angle boundaries (HABs) with misorientation angles > 15°, and the thin silver lines indicate low angle boundaries (LABs) with misorientation angles between 2° and 15°. Fig. 3(e) indicates the sampling positions, including surface of the welding zone (Z1), center of the welding zone (Z2), surface of the matrix zone (Z3) and center of the matrix zone (Z4). Fig. 3(a) and (c) shows that both of the surfaces Z1 and Z3 contain abnormal coarse grains, and the width of these grains is even larger than 200 μm. Moreover, a small amount of fine equiaxed grains can also be observed, especially in Z3. In contrast, Z2 exhibits a complete fine grain structure with an average grain size of 4.19 μm, while a bimodal grain structure formed in Z4. Additionally, many LABs remained in Z2 and Z4, which is evidence of continuous dynamic recrystallization (CDRX).

As aforementioned, PCG structure formed on the surfaces of both the welding zone (Z1) and matrix zone (Z3), while complete DRX occurred in the center of the welding zone (Z2), and partial DRX occurred in the center of the matrix zone (Z4). To further interpret this phenomenon, the distribution of strain and strain rate on the surface and in the center of the extruded P1 position are shown in Fig. 4. As is seen, Z2 exhibits the highest strain during extrusion, while relatively low values of strain were observed in Z1 and Z4. In the case of strain rate, a high value greater than 10 s^-1 was observed in Z1 and Z3, while the values of strain rate in Z2 and Z4 are in the range of 3‒5 s^-1. It was reported that a higher strain and lower strain rate are favorable for the recrystallization of Mg alloys [19,20]. Thus, a low strain and sufficiently high strain rate on the surfaces of Z1 and Z3 led to the lowest degree of DRX. Z2, located at the center of the welding zone, was subjected to a high strain and low strain rate, led to the occurrence of complete DRX. The material of Z4 experienced a lower strain and strain rate, and only partial DRX occurred. Although the DRX degree of the surface area is low, the PCG structure shown in Fig. 3(a) and (c) should not form during the extrusion process, since DAGG can only occur when the strain rate is low [9]. The air cooling process should be the dominant factor influencing the formation of PCG structure, and this mechanism is explained in the following section.

Pole figures and inverse pole figures of Z1-Z4 are plotted in Fig. 5. Z1 and Z3 exhibit a strong {0001} <11-20> texture, which agrees with the results shown in Fig. 3(a) and (c). However, a particularly intense {0001} <10-10> texture and weak {10-10} <11-20> textures are observed in Z2, and mixtures of {0001} <11-20> and {0001} <10-10> textures appear in Z4. Moreover, most of coarse grains have an orientation of <11-20> in parallel to ED, while the orientation of fine grains is <10-10> in parallel to ED. During the initial stage of porthole die extrusion, a severe shearing deformation was generated, resulting in the basal plane being parallel to the shearing plane with an intersection angle. Additionally, high compression stress along the TD during welding stage caused the rotation of basal plane, making the basal plane perpendicular to the TD [21]. Thus, the basal pole of grains in each position tends to be parallel to the TD. The tensile stress was applied along the ED during the porthole die extrusion, and the <11-20> direction is aligned along the tensile stress. On the other hand, the <11-20> oriented grains preferentially grew into deformation regions due to the large misorientation between the <11-20> oriented grains and the deformed regions, and would become coarser [22]. Then, further grain growth was achieved in the coarse <11-20> oriented grains by consuming some fine grains, and coarse grains with orientations of <11-20 > //ED formed. Consequently, a {0001} <11-20> texture was obtained, which is a typical texture in the extruded profiles. However, the {0001} <11-20> orientation was unstable, and it could rotate 30° around the c-axis by means of {10-10} <1-210> prismatic slip and transformed to the {0001} <10-10> orientation under a high strain condition [23]. Moreover, prismatic slip occurred during extrusion due to the reduced critical resolved shear stress (CRSS) at high temperatures, and the combination of basal and prismatic slip could promote the formation of the {0001} <10-10> texture. Additionally, the occurrence of recrystallization also promoted the variation of the grain orientation, which will be discussed in the next section. Thus, the highest strain and highest DRX degree were realized in Z2, which generated a strong {0001} <10-10> texture, while Z1 and Z3 had a low strain and DRX degree and consisted of strong {0001} <11-20> textures.

To investigate the dynamic precipitation behavior of the extruded P1 position, the Z1 and Z2 zones were selected for TEM analysis, and the bright field TEM images are shown in Fig. 6. It is observed that the disk-like and near-spherical precipitates, marked by red arrows, were the dominant structures in all zones. There was also a small quantity of rod-like precipitates, as marked by the blue arrows. Additionally, more precipitates were observed in the grain interior and the grain boundaries of Z2 than of Z1. It has been reported that the main strengthening precipitates in Mg-Zn alloys are β'₁ and β'₂ [[24], [25], [26]]. Under a common condition, β'₁ formed with a rod shape, and its long axis is parallel to the [0001] direction of the α-Mg matrix, while β'₂ formed with a disk shape lying on the (0001) basal plane [27]. According to previous studies [28,29], precipitates with near-spherical shapes should also be β'₂, which formed due to the shearing that occurred during extrusion process. The dynamic precipitates that formed during the high strain rate deformation were different from those that formed in static aging process. The rod-like, disk-like and near-spherical precipitates formed in the matrix at the initial stage of deformation, and their amounts increased as deformation proceeded. As abovementioned, Z2 experienced a much higher strain and a lower strain rate than the surface area of Z1. The high strain produced a large stored deformation energy, which could enhance the driving force for the nucleation of precipitates. In addition, the low strain rate provided sufficient time for the precipitates to further grow. The abovementioned two reasons should be responsible for the fact that more precipitates were generated in the center of the profile. Moreover, it is observed that there are many precipitates located on the grain boundaries of Z2, as shown in Fig. 6(d). The pinning effect on the grain growth provided by those precipitates should also contribute to the fine grain structure of Z2.

Since only partial DRX took place on the surface of the profile due to the relatively low strain and high strain rate, the surface extruded from the die bearing should have a bimodal structure that consists of many coarse grains and few fine recrystallized grains. In this study, the extruded profiles experienced an air cooling process, and some changes might occur in microstructure, especially for the surface area. Macroscopic examination of the cross-section of P1-P5 is shown in Fig. 7. It can be observed that no visible weld seams were observed in any of the profiles, which indicates that sound solid bonding was achieved due to the high temperature and high extrusion ratio [30,31]. Moreover, an inhomogeneous microstructure was observed and can be divided into two parts, viz., a surface coarse grain (SCG) zone and a central fine grain (CFG) zone. Some abnormal coarse grains were found in the SCG zone, whose amounts decreased from P1 to P5. However, although P5 contains a thick SCG zone, the size of the coarse grain is much smaller. Additionally, the thickness of the CFG zone first increases from P1 to P2 and then decreases from P3 to P5. These phenomena indicate that the extruded profiles at different positions experienced varied microstructure evolution during the air cooling process.

To further investigate the evolution of the surface microstructure, EBSD was employed in SCG zones of P2‒P5. It is found that SCG zones of all profiles consist of coarse grains and some amount of fine equiaxed grains. Moreover, the SCG zones of P1 (Fig. 3(a) and (c)) and P2 are almost entirely composed of abnormal coarse grains, while the width of the coarse grains was reduced, and a large amount of fine grains appeared in P3, P4 and P5. It is known that the final microstructure of the profiles after air cooling is determined by SRX and the grain growth behavior. That is, the effects of grain growth are more prominent than that of SRX in P1 and P2, resulting in the fact that abnormal coarse grains dominate in their SCG zones. However, the effect of SRX became stronger in P3-P5, leading to the formation of a bimodal grain structure. It was reported that the stored deformation energy, second phase particles and texture significantly affected the SRX and grain growth in Mg alloys [32]. Therefore, to prove the abovementioned inference, the stored deformation energy, texture and precipitates were investigated by numerical simulations, EBSD and TEM.

Stored deformation energy is accumulated in the form of lattice defects, such as dislocations, and will release during the air cooling process, which can facilitate the occurrence of static recovery (SRV) and SRX. The dominant softening mechanism of Mg alloys is recrystallization, since Mg alloys belong to low stacking fault energy metal. When the extrusion process is complete, the SRX process continues due to the effect of residual heat, and more fine grains will be produced. It is obvious from Fig. 3, Fig. 8 that the amount of fine grains that formed during air cooling increases from P1 to P5. As abovementioned, the driving force for recrystallization is the stored deformation energy, which is linearly related to the square of the flow stress [33]. Hence, the flow stress induced during the extrusion process was extracted based on numerical simulations and plotted in Fig. 9(a), which implies that flow stress gradually increases from P1 to P5. It implies that P1 has the lowest stored deformation energy, which is similar to that of P2, and P5 has the highest stored deformation energy. To verify the abovementioned results, the distribution of local misorientation was plotted in Fig. 9(b-f). The local misorientation is determined by calculating the average misorientation between every pixel and its surrounding pixels and assigning the mean value to that pixel, in which a higher value of local misorientation represents a higher stored deformation energy. The results show that the local misorientation for P1 and P2 are almost in the range of 0-1.5 and have average values of 0.6 and 0.62, respectively, and P5 has the highest average value of local misorientation, which agrees with the results obtained from the flow stress. Consequently, the highest stored deformation energy was produced in P5 after extrusion, where a maximum number of fine grains were obtained by SRX.

Two areas were selected from the SCG zone of P3 to make a detailed analysis of the SRX process during air cooling. As shown in Fig. 10(a), the parent grain (P-g1) was surrounded by some fine recrystallized grains (R-g1) and sub-grains (Sub-g1). The value of local misorientation shown in Fig. 10(b) is low in the parent and recrystallized grains, while it is high in the sub-grains, which implies that the occurrence of SRX consumed the stored deformation energy. Moreover, these sub-grains can become new recrystallized grains by continuing to consume the stored deformation energy. Fig. 10(d) shows another nucleation event that occurred during the recrystallization process. Several recrystallized grains (R-g2) and a sub-grain (Sub-g2) were observed inside the parent grain (P-g2). Combined with Fig. 10(d) and (e), SRX preferred to occur in an area with a high stored deformed energy until it transformed into a recrystallized grain. Therefore, a two-stage nucleation recrystallization process was observed during the air cooling process. The initial nucleation occurred in the vicinity of grain boundaries with high stored deformation energy. Then, recrystallization proceeded in the regions with a low stored energy, such as the interior of the deformed grains.

Fig. 11 shows pole figures and inverse pole figures of P2‒P5. It is seen that P2 consists of a strong {0001} <11-20> texture, which is consistent with P1. However, a mixture of a strong {0001} <11-20> texture and a {0001} <10-10> texture was obtained in P3. As the number of fine grains increased from P1‒P5, the intensity of the {0001} <11-20> texture decreased, while that of the {0001} <10-10> texture was enhanced. Fig. 10(c) and (f) shows that the occurrence of SRX promoted the transformation of the orientation from <11-20> to nearly the <10-10> direction, independent of whether SRX occurred during the first or second stage of nucleation. According to the previous study [34], the grains with the orientations of {0001} <11-20> and {0001} <10-10> are related by a rotation of 30° about <0001 > , and grain boundaries have a high mobility. Besides, it has been reported that the grains with an orientation of {0001} <11-20> grow more rapidly than those with {0001} <10-10> [35]. Thus, the grains with an orientation of {0001} <11-20> grow preferentially at the expense of {0001} <10-10> grains. Meanwhile, the {0001} <11-20> grains can also merge to reduce the interfacial energy. Therefore, the degree of SRX in P1 and P2 is low due to the low stored deformation energy, and grain growth dominated during the air cooling process. Consequently, the microstructure of P1 and P2 consists of abnormal coarse grains with orientations of {0001} <11-20 > . In the case of P3‒P5, more recrystallized grains were produced due to the high stored deformation energy, which led to the formation of some <10-10> oriented grains. Finally, a <10-10> - <11-20> double fiber texture parallel to the ED formed.

Second phase particles also play an important role in SRX and grain growth, and play different roles with different sizes. Coarse particles can promote the nucleation of recrystallization, which is known as particle stimulated nucleation (PSN), while fine particles can retard the mobility of dislocations and grain boundaries, which is known as a pinning effect [36]. In this study, both coarse and fine particles were observed. Fig. 12 shows the SEM images and corresponding EDS maps of second phase particles in the SCG zone of P4. A large amount of Zr-rich particles was found in the fine grain area, while particles did not appear in the coarse grain area. The existence of Zr can promote the nucleation of SRX. Moreover, the precipitates that formed after extrusion and air cooling were examined by TEM. The bright field TEM images of P4 and P5 are shown in Fig. 13. The precipitates contain three forms: rod-like, disk-like and near spherical shapes, where the disk-like and near-spherical precipitates are marked by red arrows, and the rod-like precipitates are marked by blue arrows. The majority of the precipitates in P4 and P5 are disk-like and near spherical, and several rod-like precipitates also exist. Moreover, the amount of precipitates in P5 is much more than that in P4. However, in comparison with Fig. 6, the amount of precipitates in P4 and P5 is much more than that in P1. As abovementioned, the rod-like precipitates are β'₁ phase, while the disk-like and near spherical precipitates are β'₂ phase. It was reported that the precipitates that formed during the static processing of Mg-Zn alloys follows the sequence of a supersaturated solid solution (SSSS) → Guinier-Preston (GP) zones → β'₁ → β'₂ → β equilibrium phase [37]. From Fig. 6, Fig. 13, it can be concluded that the amount of precipitates tends to increase from P1 to P5, which can be attributed to the different amounts of stored deformation energy. P5 has the highest stored deformation energy, and the second phase particles more easily nucleated and precipitated. Moreover, many fine particles precipitated on the grain boundaries, which can impede grain growth due to the pinning effect. The high resolution TEM image of the rod-like and disk-like precipitates, the corresponding fast Fourier transform (FFT) pattern and the inverse FFT images are shown in Fig. 14. As is seen, the rod-like and disk-like precipitates are nearly coherent with the Mg matrix. The orientation relationship between β'₁ and Mg is (2-1-10)_β'1//(0001)_Mg and (01-1-1)_β'1//(0-110)_Mg, while that of β'₂ and Mg is (01-10)_β'2//(0001)_Mg and (0001)_β'2//(0-110)_Mg. The TEM results agree well with the previously reported study [38].

Based on abovementioned discussion, the microstructure evolution of profile surface during the porthole die extrusion and subsequent air cooling can be summarized. During the extrusion process, the materials on the surface experienced a higher strain rate and lower strain, resulting in the occurrence of partial DRX. Consequently, at the moment when the profile was extruded out from the die bearing, the microstructure consisted of a majority of coarse grains with an orientation of <11-20> in parallel to ED and a few DRXed grains. Then, during air cooling, a two stage SRX process took place, and a large amount of fine recrystallized grains with an orientation of <10-10 > in parallel to ED formed. In addition, the coarse grains with an orientation of <0001> {11-20} tend to grow at the expense of grains with other orientations. The final microstructure is the result of the competition between SRX and grain growth. Due to the low stored deformation energy and few precipitates, grain growth is more prominent than SRX in P1 and P2. Consequently, the SCG zones of P1 and P2 are almost entirely composed of abnormal coarse grains. However, from P3 to P5, the stored deformation energy and the amount of precipitates gradually increased, by which the SRX was enhanced and the grain growth was restricted. Hence, the size of the coarse grains decreased, and the amount of fine grains increased in the SCG zones of P3, P4 and P5. The stored deformation energy is the factor dominating the surface microstructure through effective regulation over the nucleation of recrystallization and precipitate. High stored deformation energy could form a structure that consisted of relatively coarse grains and a large amount of fine grains. In contrast, low stored deformation energy suppressed formation of SRX and precipitates, consequently resulting in a surface microstructure with abnormal coarse grains.

The engineering stress‒engineering strain curves, tensile strength and elongation of P1, P2, P3, P4 and P5 are depicted in Fig. 15. For all profiles, strain hardening rate first increases and then decrease with the increase in the strain. P1 and P4 exhibit almost the same and lowest tensile strength of approximately 279 MPa. P2 has the highest tensile strength of 293 MPa and a high elongation of 16.5%, and both values are similar to those of P3. P5 achieves the best elongation of 16.7% and a relatively high tensile strength of 287 MPa. At ambient temperature, only the basal slip systems are active due to their low CRSS [39]. However, as aforementioned, whether in the center or on the surface of profile, the basal plane of most grains is perpendicular to the TD, which is also perpendicular to the tensile direction. Thus, it was difficult for basal slip to occur, resulting in the increase in the hardening rate. With the increase in the strain, prismatic slip and twinning occurred, which led to the decrease in the hardening rate [40]. As shown in Fig. 7, the thickness of the CFG zone in P1 and P4 is relatively thinner, which is one reason for the low tensile strength and elongation. Contrarily, the thicker CFG zones in P2 and P3 contributed to their high tensile strength and elongation. Although P5 has the thinnest CFG zone, the grains in the SCG zone had been greatly refined, and the abnormal coarse grains were absent, as shown in Figs. 7(e) and 8 (d). It is reported that the capacity of fine and coarse grains to accommodate dislocations generated by the deformation process is different, and dislocations are easy to accumulate and are blocked at the grain boundaries. Thus, it is difficult for dislocations to slip, and the work-hardening ability is improved, which can help delay localized deformation under tensile test, and therefore, elongation increases [41]. As a result, P5 possesses the highest elongation.

Fig. 16 shows that all profiles except P2 fractured at the position of the welding zone. Generally, the fracture tends to occur in the area near the weld seam, which is regarded as the weakest zone [42]. In this study, P2 fractured on the matrix zone, which means that a sufficient strength had been achieved on the weld seam. Fig. 7(b) shows that the thickest CFG zone was achieved in the welding zone of P2, while a bimodal grain structure was observed in the matrix zone. Thus, the welding zone of P2 has a relatively high tensile strength, and the sample fractured in the matrix zone. To further investigate the fracture mechanism of the extruded profiles, the fracture morphologies were observed using SEM, and the results are presented in Fig. 17. In P1, P3 and P4, the center of the fracture surface is composed of a large amount of small dimples and several large dimples, while many cleavage planes, microcracks, tear ridges and few dimples are found on the surface areas. However, in P2 and P5, the cleavage planes and dimples are located at the center of the fracture surface, and many cleavage planes as well as some microcracks, tear ridges and dimples are observed on the surface areas. Therefore, the fracture of all the profiles follows two modes: the ductile fracture in the center of the profiles and transgranular fracture on the surface of the profiles [43,44]. It was reported that the coarse grains displayed a poor plastic deformation capacity. During the tensile test, twinning easily occurred with the increase in the strain, especially in the SCG zone. The microcracks tended to form along the twinning regions [45], and a transgranular fracture formed with the propagation of microcracks. Fig. 7(a) and (d) shows that P1 and P4 have the thinnest CFG zone, and they possess the worst mechanical properties. Moreover, although P5 has the thinnest fine grain layer, the abnormal coarse grains on its surface have been replaced by a mixture of coarse grains and fine grains, as shown in Fig. 7(e). Thus, the existence of abnormal coarse grains on the surface greatly deteriorated the strength and elongation.

Fig. 18 plots the hardness distribution on the cross-section of the extruded profiles. The detailed positions for the hardness tests are presented in Fig. 18(f). Three tests were repeated at each position, and the average value was calculated. It can be seen that hardness gradually increases from the surface to the center of the profile. The area with the high value of hardness is consistent with the CFG zone, as shown in Fig. 7. Coarse or abnormal coarse grains can be found in the SCG zones of all the profiles, which is the reason for the low hardness on the profile surface, while the fine grain structure in the CFG zones contributes to the enhancement of the hardness. Additionally, more particles precipitated in the CFG zones than in the SCG zones, which also affect the hardness distribution. Moreover, the surface hardness first increased from P1 to P4, and then decreased in P5. The amount of fine grains increased from P1 to P4, and the amount of precipitates also increased, which caused the increase in the hardness. However, obvious grain growth is observed for P5. This is the reason for the decrease in the hardness.

The wear rates obtained from the dry sliding wear test are shown in Fig. 19. From P1 to P5, the wear rate gradually decreases, which means that the wear resistance increases. During the wear test, plastic deformation took place on the surface and subsurface of the profile due to the repeated action of the applied loads. Such a deformation is realized through dislocation slip, resulting in the formation of microcracks. The cracks will propagate on the surface, and some material debris will fall off of the surface, which means that wear will occur [46]. The SCG zone of P1 consisted of complete abnormal coarse grains, which exhibit a poor plastic deformation capacity and the worst wear resistance [47]. From P1 to P5, the amount of abnormal coarse grains was reduced, and more fine grains and precipitates formed in the SCG zone, resulting in the better wear resistance. In the case of P5, the fine grains become coarse through grain growth, but the homogeneity of the microstructure is improved, which is beneficial to the wear resistance. Therefore, the wear rate of P5 is slightly lower than that of P4. Fig. 20 shows the SEM images of the worn surfaces after the dry sliding tests. Fig. 20 shows that grooves with wear debris were found on the worn surface of each sample, which were caused by the movement of the steel ball on the surface of the softer samples and indicates the occurrence of abrasive wear [48]. Moreover, some black particles attached to the worn surface were observed in P1, P2, P3 and P4, as indicated by the red arrows in Fig. 20, which is evidence of oxidation abrasion. However, few black particles were found on the worn surface of P5. Therefore, the wear mechanisms of the friction process for P1, P2, P3 and P4 are abrasive wear and oxidation abrasion, while that of P5 is abrasive wear.

Porthole die extrusion was conducted to investigate the formation and evolution of surface coarse grains during extrusion and the subsequent air cooling process. The effects of surfaces with different microstructures on the tensile properties, hardness and wear properties were also studied. The conclusions are as follows.

(1) Inhomogeneous microstructure formed on the cross-section of the extruded profiles. The surface of the profile mainly consisted of abnormal coarse grains. The center of the welding zone exhibited a fine grain structure with an average size of 4.19 μm, while a bimodal grain structure formed in the center of the matrix zone. Most of the abnormal coarse grains had orientations of <11-20> in parallel to ED, while the orientation of the fine grains was <10-10> in parallel to ED. Disk-like, near-spherical and rod-like precipitates were observed in all the zones, while their amounts are much less on the profile surface than in the profile center.

(2) The final surface microstructure depends on the competition between static recrystallization and grain growth. P1 and P2 with low stored deformation energy suppressed formation of precipitates and consequently facilitated grain growth rather than recrystallization, resulting in the formation of abnormal coarse grains. With increasing the stored deformation energy from P3 to P5, the amount of precipitates and SRXed grains gradually increased, resulting in the retardation of grain growth. Hence, the size of the coarse grains decreased, and the amount of fine grains increased in the SCG zones of P3, P4 and P5. The stored deformation energy was the factor dominating the surface structure through effective regulation over nucleation of the precipitates and recrystallization.

(3) P4 had the thickest CFG zone and exhibited the best tensile property, while P1 and P3 had the thinnest CFG zone and the poorest tensile property. From the surface to the center of the profile, the hardness gradually increased due to the variation of the grain size and the amount of precipitates. The wear rate gradually decreased from P1 to P5. Abrasive wear and oxidation abrasion were found in P1 to P4, while only abrasive wear appeared in P5.

This work was financially supported by the National Natural Science Foundation of China (No. 51875317), the Development Program of Shandong Province (No. 2019GGX104087), and the National Natural Science Foundation of Shandong Province (No. ZR2019QEE030).