Microstructure and Mechanical Properties of ARB Processed Aluminium with Different Purities

Abstract

99.2% Al (2N-Al), 99.99%Al (4N-Al) and 99.999%Al (5N-Al) were deformed to high strains by accumulative roll-bonding (ARB) at room temperature, and microstructure and mechanical properties were systematically characterized. During the ARB process, original coarse grains were subdivided by deformation-induced high-angle boundaries into nano-scale grains, where redundant shear strain introduced in the near-surface layers by friction in the non-lubricated rolling significantly accelerated the formation of nanostructures. It was found that spacing and fraction of high-angle boundaries can be explained by total equivalent strain taking the effect of shear strain into account. Quite uniform nanostructures dominated by high-angle boundaries were obtained after 6 cycles of ARB for 2N-Al and 4N-Al, but the boundary spacing was smaller in the 2N-Al than in the 4N-Al. On the other hand, in the case of 5N-Al, recrystallization and grain growth occurred during the roll-bonding process, and nanostructures were not able to be obtained. It was suggested that an increase in the amount of impurities is effective to increase the stability of the nanostructures and to randomize the deformation texture leading to a high fraction of high-angle boundaries.

1. Introduction

Nanostructured metals with average grain sizes smaller than 1 μm are known to reveal outstanding mechanical properties, compared with conventionally coarse grained metals.¹⁾ Severe plastic deformation (SPD) process is a promising strategy to produce bulk nanostructured metals having large dimensions. One dominant mechanism for the formation of nanostructures during SPD is grain subdivision.^2,3) Namely, during the process, original coarse grains are subdivided by deformation-induced boundaries into nano-scale, and quite uniform nanostructures dominated by high-angle boundaries are formed after ultrahigh strains. However, it has also been reported^4–8) that grain size or boundary spacing of nanostructures tend to be saturated after ultrahigh strains. This observation indicates that not only deformation-induced grain subdivision but restoration processes, i.e. recovery and boundary migration, during deformation play a significant role to determine the nanostructures produced by SPD.

Factors that may affect the microstructural evolution during SPD can be classified into two categories: materials factors and deformation factors. Materials factors include solute atom, second phase, initial grain size, stacking fault energy, etc., while deformation factors include strain, strain rate, deformation temperature, deformation mode, strain path, etc. To understand the overall mechanism of nanostructure formation during SPD furthermore, effects of these factors should systematically be studied.

The present study investigates the impurity effect on microstructure evolution during SPD in pure aluminium. Three kinds of pure aluminium with different levels of purity are deformed to high strains by accumulative roll-bonding (ARB),⁹⁾ a kind of SPD process to produce bulk nanostructured metals, and deformation structures are characterized in detail. Some results on mechanical properties are also demonstrated in the paper. In the discussion, correlation between plastic strain introduced by ARB process and microstructure evolution is studied.

2. Experimental

2.1 ARB process

A commercial purity aluminium JIS A1100 (99.2 mass% pure), a high purity aluminium (99.99%) and an ultra-high purity aluminium (99.999%) were used in the present study, which are hereafter referred to as 2N-Al, 4N-Al and 5N-Al, respectively. The main impurities are 0.12% of Si, 0.51% of Fe and 0.11% of Cu for 2N-Al, and 0.003% of Si, 0.003% of Fe and 0.001% of Cu for 4N-Al. For 5N-Al, 2.8 mass ppm of Cu is included and other elements are all less than 1 mass ppm. Fully-recrystallized samples were used as starting materials for all alminiums. The average grain sizes of the initial fully recrystallized structures were 18 μm for 2N-Al, 41 μm for 4N-Al and 84 μm for 5N-Al.

Sheets with a thickness of 1 mm, a width of 40 mm and a length of 250 mm were prepared from the starting materials. The following procedure of ARB was applied for three kinds of the materials. After degreasing and wire-brushing contact surfaces, two sheets were stacked to be 2 mm thick and roll-bonded at room temperature by 50% thickness reduction in one pass without using any lubricant. A two-high mill with rolls 310 mm in diameter was used, and the roll-bonding was carried out at a rolling speed of 17.5 m min⁻¹. The average strain rate during rolling can be estimated to be 19 s⁻¹, where this strain rate is calculated only from the thickness reduction in one-pass rolling. The roll-bonded sheets were water-cooled immediately after the rolling to minimize structural changes caused by deformation-induced heat generation. A procedure of cutting, stacking and roll-bonding is referred to as 1 cycle of ARB and was repeated up to 6 cycles without changing the rolling direction (RD). The ARB samples were always stored in a freezer kept at −50℃ till they were subject to experiments, to avoid unnecessary recovery or recrystallization during the storage.

2.2 EBSD measurement

In this experiment, ARB process has been done under a non-lubricated condition. In this condition, due to the friction between the rolls and sheet surfaces, a large amount of redundant shear strain is introduced in the surface layers.^10,11) By repetition of stacking and roll-bonding in the ARB process, severely-sheared layers on the surface come into the center layers in each ARB cycle, leading to a quite complicated shear strain distribution throughout the thickness of the samples.^10–12) Therefore, deformation structure may significantly vary depending on the thickness location as well as the number of ARB cycles, as previously reported in the case of ARB processed interstitial free (IF) steel.¹¹⁾ In this study, electron backscatter diffraction (EBSD) measurements were carried out at different thickness locations for the ARB samples to observe the microstructural heterogeneity through the sheet thickness.

Longitudinal sections perpendicular to the transverse direction (TD), including the normal direction (ND) and RD, of the samples were mechanically polished with SiC emery papers and then electrically polished in a solution of 30 vol% nitric acid (HNO₃)+70 vol% ethanol (C₂H₅OH) at −15℃ at a voltage of 20 V. EBSD measurements were carried out at different thickness locations. In this paper, the thickness location is indicated by a parameter t/t₀, where t is the distance from the thickness center to the central position of the EBSD scanned area and t₀ is the thickness of the sheet. For instance, the location of t/t₀ = 0 and t/t₀ = 0.5 corresponds to the thickness center and the surface, respectively. EBSD measurements were carried out using a software TSL OIM Data Correction in a Philips XL 30S field emission SEM operated at 15 kV. The obtained EBSD data were analyzed using a software TSL OIM Analysis. Boundaries with misorientation angles less than 2° were ignored in the analysis to remove errors in the determination of such very low-angle boundaries in EBSD measurements. Boundary spacing was determined from the EBSD maps by intercept method.

2.3 TEM observation

Deformation microstructures of the ARB processed samples were also observed by transmission electron microscopy (TEM). Thin foils parallel to the TD plane of the samples were prepared by mechanical polishing and electrical polishing using a similar method for EBSD sample preparation. TEM observations were carried out in a Hitachi H-800 TEM operated at 200 kV. Also for TEM images, boundary spacing was determined by intercept method, where all boundaries observed in the images were taken into account for the determination of the spacing. Dislocation density was measured by Ham's method¹³⁾ using TEM images taken under multi-beam diffraction conditions to reveal all dislocations.

2.4 Tensile test

Uniaxial tensile tests were carried out at room temperature. Tensile specimens with a gauge length of 10 mm, width of 5 mm and thickness of 1 mm were prepared from the ARB processed samples, with the tensile axis parallel to the RD. The tests were carried out at a constant crosshead speed of 0.5 mm min⁻¹, corresponding to an initial strain rate of 8.3 × 10⁻⁴ s⁻¹. Tensile elongation was measured by a clip-on extensometer.

3. Results

3.1 EBSD measurements

3.1.1 Commercial purity aluminium (99.2% Al: 2N-Al)

Figure 1 shows several examples of grain boundary maps obtained from the EBSD measurements for the ARB processed samples of 2N-Al. Results from different thickness locations, i.e. near center (t/t₀ ∼ 0), quarter (t/t₀ ∼ 0.2) and surface (t/t₀ ∼ 0.5), are shown. Green and red lines indicate high-angle boundaries with misorientation angles above 15° and low-angle boundaries of 2–15°, respectively. In the 1-cycle ARB sample (Fig. 1(a)), the center and quarter layers are dominated by low-angle boundaries, even though some of high-angle boundaries of initial grains are observed. However, a large number of high-angle boundaries are already introduced near the surface layers even after 1 cycle of ARB, indicating that the redundant shear strain caused by non-lubricated rolling accelerates the formation of deformation-induced high-angle boundaries. The 2-cycle ARB sample (Fig. 1(b)) has layers dominated by high-angle boundaries in the near-center layer as well as the near-surface layer. This is quite reasonable since the surface layer severely sheared in the first ARB cycle came to the thickness center in the second ARB cycle. The distribution of severely sheared layers is expected to become more homogeneous with increasing the number of ARB cycles, leading to a more uniform distribution of high-angle boundaries through the thickness (Fig. 1(c)). Finally, 6 cycles of ARB (Fig. 1(d)) produces an elongated ultrafine structure dominated by high-angle boundaries uniformly throughout the sample thickness. These observations are in good agreement with the results of the IF steel samples deformed by ARB without lubrication.¹¹⁾

Fig. 1

Boundary maps at various thickness locations in the 2N-Al ARB processed by various cycles. The thickness location of the measured area is indicated in each figure.

From the EBSD results obtained, spacing of high-angle boundaries along the ND (d_t,HAGB) and fraction of high-angle boundaries (f_HAGB) were determined, and plotted as a function of thickness location in Fig. 2. The high-angle boundary spacing (Fig. 2(a)) in the 1-cycle ARB sample is large in the near-center layer but decreases with approaching the surface, due to the effect of the redundant shear strain. With increasing the number of ARB cycles, the spacing of high-angle boundaries decreases, depending on the expected distribution of shear strain, but it tends to be saturated at high strains. A quite uniform distribution of high-angle boundary spacing (200–300 nm) is observed through the thickness in the 6-cycle ARB sample. The fraction of high-angle boundaries (Fig. 2(b)) is quite low at the thickness center of the 1-cycle ARB, but it becomes larger with approaching the surface. The fraction of high-angle boundaries increases with increasing the number of ARB cycles, which also corresponds well with the expected accumulation of shear and rolling strain. The distribution of the fraction tends to become more homogeneous with increasing the number of ARB cycles and saturated after several number of ARB cycles, as was in the distribution of high-angle boundary spacing. Approximately 80% of high-angle boundary fraction is observed throughout thickness of the 6-cycle ARB sample sheet. These tendencies in the structural parameters are also quite similar to those observed in IF steel samples deformed by non-lubricated ARB.¹¹⁾

Fig. 2

Average spacing of high-angle grain boundaries (a) and fraction of high-angle grain boundaries (b) through thickness of the 2N-Al ARB processed by various cycles.

3.1.2 High purity aluminium (99.99% Al; 4N-Al)

Similar EBSD measurements have been applied for 4N-Al ARB samples, and results are shown in Figs. 3 and 4. Microstructural evolution through the sheet thickness for the 4N-Al samples is qualitatively similar to those for 2N-Al shown in Figs. 1 and 2. Deformation-induced high-angle boundaries are preferentially introduced in the severely sheared surface layers after 1 cycle (Fig. 3(a)), and microstructures becomes more and more homogeneous with increasing the number of ARB cycles (Fig. 3(b)–(c)). A quite uniform structure dominated by high-angle boundaries can be obtained after 6-cycle ARB (Fig. 3(d)), as was previously reported in the reference¹²⁾. It should, however, be emphasized here that some specific differences are recognized between the results of 2N-Al and 4N-Al. The structures dominated by high-angle boundaries after several ARB cycles are not elongated but quite equiaxed (Fig. 3) in the 4N-Al, compared with that in the 2N-Al. The minimum average spacing of high-angle boundaries (Fig. 4(a)) is approximately 1 μm in the 4N-Al, which is much larger than that in the 2N-Al (200–300 nm). These differences suggest that boundary migration occurred more significantly in the 4N-Al during the ARB process due to the higher purity. Another interesting difference is that the fraction of high-angle boundaries in the 4N-Al ARB sample is about 70% at the maximum (Fig. 4(b)), which is smaller than that in the 2N-Al ARB sample (～80%). It can therefore be concluded that an increase in the impurities of aluminium is more effective to obtain smaller-sized nanostructures with a higher fraction of high-angle boundaries.

Fig. 3

Boundary maps at various thickness locations in the 4N-Al ARB processed by various cycles. The thickness location of the measured area is indicated in each figure.

Fig. 4

Average spacing of high-angle grain boundaries (a) and fraction of high-angle grain boundaries (b) through thickness of the 4N-Al ARB processed by various cycles.

3.1.3 Ultra-high purity aluminium (99.999% Al; 5N-Al)

Microstructural distributions of ARB samples were also investigated for the 5N-Al. Figure 5 shows through-thickness EBSD maps of the 5N-Al samples ARB processed by various cycles, where image quality (IQ) maps and grain boundary (GB) maps are indicated. It is clearly seen that the microstructural distribution are quite similar in the 5N-Al samples regardless of the number of ARB cycles. In the near-center layers, typical deformation structures dominated by low-angle boundaries are observed, while in the layers from the surface to the quarter thickness, recrystallized grains surrounded by high-angle boundaries are dominant. Some recrystallized grains are also present in the center layers. It should be noted that a lot of low-angle boundaries are observed within such recrystallized grains, indicating that these recrystallized grains have been formed mainly by dynamic recrystallization during the rolling. This can easily be understood due to the ultra-high purity of the material, but it is somewhat interesting that dynamically recrystallized grains have been observed near the surface layers preferentially.

Fig. 5

EBSD maps through thickness of the 5N-Al ARB processed by (a) 1cycle, (b) 2cycles, (c) 4cycles and (d) 6cycles. Image Quality (IQ) maps and grain boundary (GB) maps of the same areas are shown.

This could be understood in the following way. The non-lubricated rolling used in this study led to a large amount of redundant shear strain near the surface layer, introducing a high density of both dislocations and deformation-induced high-angle grain boundaries. Such high density of high-angle boundaries should act as nuclei for recrystallization, and the high-density of dislocations should act as driving force for recrystallization. A large amount of redundant shear strain should also lead to a large plastic work in the surface layers, providing a large amount of heat generation. Therefore, dynamic recrystallization may occur during rolling much more easily in the surface layers than in the center layers. Since the dynamically recrystallized grains in the surface layers are moved to the thickness center in the next ARB cycle and experience a conventional rolling deformation, the obtained structure in the center layer will be a typical deformation structure without producing any ultrafine structure. In other words, for such an ultra-high purity aluminium, plastic strains applied cannot be accumulated but can easily be released by dynamic recrystallization, so that it is difficult to obtain ultrafine grained structures in the ultra-high purity aluminium.

3.2 TEM observation

To reveal more detailed features of the obtained deformation structures, TEM microstructures were observed for the 6-cycle ARB samples of the 2N-Al, 4N-Al and 5N-Al. Figure 6 indicates TEM images obtained from near center layers. The 2N-Al ARB sample (Fig. 6(a)) shows a quite uniform ultrafine lamellar structure elongated to RD. The average spacing of the lamellar boundaries and the dislocation density between the boundaries were determined to be 180 nm and 1.3 × 10¹⁴ m⁻², respectively.¹⁴⁾ In the 4N-Al (Fig. 6(b)), more equiaxed structure is obtained. The average boundary spacing is 690 nm,¹⁵⁾ which is much larger than the boundary spacing of the 2N-Al ARB sample. The dislocation density in the 4N-Al sample was measured to be 1.2 × 10¹³ m⁻²,¹⁵⁾ much lower than that in the 2N-Al. These differences in the deformation structure between the 2N-Al and 4N-Al are caused by the enhanced recovery and boundary migration in the 4N-Al. On the other hand, typical dislocation cell structures are observed in the 5N-Al after 6-cycle ARB (Fig. 6(c)), corresponding with the observation results shown in Fig. 5.

Fig. 6

TEM microstructures of the 6-cycle ARB processed samples. (a) 2N-Al, (b) 4N-Al [15] and (c) 5N-Al.

It can be concluded from the above results that in aluminium, the purity of 99.99% or less is required to obtain uniform ultrafine grained structures dominated by high-angle boundaries by the SPD process.

3.3 Mechanical properties

Tensile tests at room temperature were carried out for the ARB processed samples. Figure 7 shows nominal stress-strain curves. The ultimate tensile strength and total elongation obtained from the curves are plotted as a function of the number of ARB cycles in Fig. 8.

Fig. 7

Nominal stress-strain curves of the pure aluminum samples ARB processed by various cycles at room temperature. (a) 2N-Al, (b) 4N-Al and (c) 5N-Al.

Fig. 8

(a) Ultimate tensile strength and (b) total elongation of the ARB processed samples.

In the 2N-Al, the strength increases gradually with increasing the number of ARB cycles, and the ultimate tensile strength reaches as high as 330 MPa after 6 cycles of ARB, which is approximately three times higher than the starting material. This tendency is reasonable considering the microstructural changes shown in Figs. 1 and 2, i.e. the formation of ultrafine lamellar structure proceeded with increasing the number of ARB cycles, and the 6-cycle ARB produced a uniform ultrafine structure dominated by high-angle boundaries. On the other hand, the total elongation significantly decreases after 1 cycle, and it tends to be almost constant after further ARB cycles. The uniform elongation is 2–3% and the total elongation is approximately 10% in each sample ARB processed. These tendencies are in good agreement with the results previously reported in a 99.2% pure aluminium deformed to high strains.¹⁶⁾

In the 4N-Al, the change in the strength and ductility is quite different from that in the 2N-Al. The strength increases with increasing the ARB cycles and the ultimate tensile strength reaches the maximum of 135 MPa in the 2-cycle ARB, but further ARB deformation leads to a gradual decrease in strength rather than strengthening. The ultimate tensile strength decreases down to 100 MPa in the 6-cycle ARB. Such deformation-induced softening during SPD has also been observed in 99.99% pure aluminium deformed to high strains, e.g. by high pressure torsion (HPT),¹⁷⁾ equal channel angular extrusion (ECAE)¹⁸⁾ and accumulative channel-die compression bonding (ACCB).¹⁹⁾ Such a softening behavior should be related to the enhanced recovery and boundary migration during the ARB process, as was observed in Figs. 3 and 6(b), which led to a decrease in both dislocation strengthening and grain boundary strengthening. The change in the ductility of the ARB processed 4N-Al is also interesting. The total elongation decreases by the 1-cycle ARB, but it starts to increase with increasing the strain after the ARB of 2 cycles and above. Improvement of ductility during severe plastic deformation has also been reported previously in SPD processed commercial purity aluminium (99.5% pure)²⁰⁾ or AA8011 aluminium alloy (98.6% pure).²¹⁾ The mechanism of such behavior is related to enhanced recovery during the tensile test due to a high density of high-angle boundaries as well as high purity of the material.^20,21)

In the stress-strain curves of the 5N-Al, the strength slightly increases by 1-cycle ARB, but it does not change so much after 2 cycles and above, keeping a relatively low strength of about 70 MPa even after high strains. The elongation of the 5N-Al ARB samples is relatively high regardless of the number of ARB cycles. This may also correspond well with the observed microstructural evolution. As was shown in Fig. 5, the 5N-Al ARB samples revealed a mixture of deformation structures and dynamically recrystallized grains, and the microstructures were quite similar independent of the number of ARB cycles. Therefore, the strength cannot be so high, even though ultrahigh strain is applied by the ARB process. It can be concluded from the above-mentioned observations that the purity of 99.99% or less should be required to obtain high-strength aluminium after the SPD.

4. Discussion

4.1 Correlation between equivalent strain and microstructural parameters

This study has demonstrated that microstructural evolution during ARB is significantly accelerated by redundant shear strain introduced in the surface layer. In the 2N-Al and 4N-Al, the formation of nanostructures by grain subdivision proceeds much faster in the near-surface layers than in the thickness-center layers of the ARB sheets. However, it is not yet clear how the redundant shear deformation affects the formation of nanostructures. In this section, we try to correlate the microstructural evolution with the redundant shear strain introduced during the ARB process.

To quantify the effect of redundant shear strain, the shear strain distribution through the thickness of the ARB sheets was evaluated by the embedded-pin method, which had originally been established by Sakai et al.²²⁾ In this method, the cylinder-shaped pin was embedded in the undeformed sheet, and the sheet with the pin was deformed by 1 cycle of ARB. The distortion of the pin after the rolling was observed by optical microscopy, and the shear strain distribution through the sheet thickness was evaluated. For the estimation of shear strain, several assumptions have been taken into account. When the non-lubricated rolling is carried out, shear deformation is first introduced at the surface towards the rolling direction before the neutral point in the roll bite, but after the neutral point the reverse shear deformation is introduced towards the opposite direction only in the near-surface region. Thus, the total shear strain was calculated by the sum of the first shear strain and the reverse shear strain estimated from the embedded-pin method. Assuming that the shear strain introduced by the first ARB cycle was repeated in the same manner after the second ARB cycle and above, the total shear strain distribution was estimated for each ARB cycle. Finally, the obtained shear strain was transferred into equivalent strain, using the von Mises equivalent strain equation. A detailed procedure of the method for estimating shear strain and equivalent strain in the ARB process has been described in the previous paper¹¹⁾.

The spacing and fraction of high-angle boundaries obtained by EBSD measurements are plotted as a function of the estimated equivalent strain in Figs. 9(a) and 9(b), respectively, for the 2N-Al and 4N-Al. It is clearly seen that both the spacing and fraction of high-angle boundaries can be well explained by a single curve in each material as a function of the total equivalent strain taking account of the redundant shear strain. It should be noted here that the non-lubricated ARB process in the present study may introduce a quite complicated rolling and shear deformation mode as a function of both the thickness location and the number of ARB cycles, due to the redundant shear strain in the surface layers in each ARB cycle, as pointed out previously. The introduction of redundant shear deformation in the surface layers during the rolling causes a very complicated deformation with simultaneous changes in strain, strain rate, deformation mode and strain path. Thus, it is somewhat surprising that the deformation microstructure can basically be explained by the total equivalent strain. This indicates that, at least in the experimental conditions of ARB process in the present study, total equivalent strain is a dominant factor to determine the formation of deformation-induced high-angle boundaries, but the changes in the deformation mode, strain rate or strain path have minor effects.

Fig. 9

Relationship between microstructural parameters and equivalent strain in the 2N-Al and 4N-Al samples ARB processed. (a) Average spacing of high-angle grain boundaries and (b) fraction of high-angle grain boundaries.

It is also seen from Fig. 9 that there is a significant difference in the spacing and fraction of high-angle boundaries between 2N-Al and 4N-Al. For both materials, the spacing of high-angle boundaries (Fig. 9(a)) decreases and the fraction of high-angle boundaries (Fig. 9(b)) increases with increasing the equivalent strain, but tends to be saturated above an equivalent strain of ～7, indicating that the grain subdivision process and the recovery and/or boundary migration process are balanced during rolling. It is seen that the obtained boundary spacing is always smaller in the 2N-Al than in the 4N-Al, and the saturated boundary spacing is also smaller in the 2N-Al (200–300 nm) than in the 4N-Al (～1 μm). These results indicate that the increase of impurities is more effective to increase the stability of ultrafine grained structures obtained by the SPD.

It should also be emphasized in Fig. 9(b) that the curve of high-angle boundary fraction for the 2N-Al tends to be higher than for the 4N-Al in all strain range, and the saturated fraction is higher in the 2N-Al (～80%) than in the 4N-Al (～70%). This result is discussed in terms of texture randomization in the next section.

4.2 Randomization of texture by addition of impurities

The fact that a higher fraction of high-angle boundaries was observed for the 2N-Al than for the 4N-Al in Fig. 9(b) may suuggest a possibility that deformation texture formed by the ARB in the 2N-Al is more randomized than in the 4N-Al. To verify this, texture components of the 2N-Al and 4N-Al is compared based on the EBSD measurements. Figure 10 shows the texture component maps in the center and surface layers of the 6-cycle ARB processed 2N-Al and 4N-Al. In these figures, high-angle and low-angle boundaries are indicated by black and gray lines, respectively, and typical rolling texture components and shear texture components in face-centered cubic metals are indicated in different colors. Components of {112}<111> (Copper), {110}<112> (Brass) and {123}<634> (S) are chosen as typical rolling texture components,²³⁾ while {001}<110> (Rotated Cube), {111}<110> and {111}<112> are chosen as typical shear texture components.²⁴⁾ Tolerance angle of 15° from each ideal orientation was used to determine the area fraction of each texture components. White-colored components that do not belong to either rolling or shear texture components are considered as random texture component.

Fig. 10

Texture component maps in the center and surface layers of the 6-cycle ARB processed 2N-Al and 4N-Al.

It can be seen from Fig. 10 that for both materials (2N-Al and 4N-Al) rolling texture components are preferentially observed in the center layers, while shear texture components are frequently observed in the surface layers. This corresponds well with the fact that the roll-bonding has been carried out under the non-lubricated condition. However, random textured grains seem to be more dominant in each area. Based on the EBSD analysis, area fraction of three rolling texture components are summed and considered as the rolling texture fraction, while sum of three shear texture components are considered as the shear texture fraction. The area fractions of rolling texture, shear texture and random texture are summarized in Table 1. As was observed from Fig. 10, the center and surface layers have a weak rolling texture or weak shear texture, respectively, in both materials, but all regions are dominated by random texture with fractions of 59–70%. This is probably because the ARB deformation under non-lubricated condition introduces a very complicated deformation mode with a mixture of conventional rolling and shear during the repetition of cutting, stacking and roll-bonding.^10,12)

Table 1 Area fractions of different texture components in the 2N-Al and 4N-Al ARB processed by 6-cycle.

	2N-Al, 6-cycle ARB		4N-Al, 6-cycle ARB
	Center	Surface	Center	Surface
Rolling texture (%)	25.9	5.9	38.1	7.5
Shear texture (%)	3.9	24.3	2.5	27.5
Random texture (%)	70.2	69.8	59.4	65.0

It should be noted that the fraction of random texture components is higher in the 2N-Al (～70%) than in the 4N-Al (～60%), demonstrating that the deformation texture in the 2N-Al is more randomized than in the 4N-Al, in good agreement with the fact that the higher fraction of high-angle boundaries is observed in the 2N-Al than in the 4N-Al as shown in Fig. 9(b). This can be understood as follows. In the 2N-Al used in this study, some amount of Si, Fe and Cu atoms are included and they exist as solute atoms or precipitates in the aluminium matrix. Such solute atoms or precipitates could suppress dislocation movement and enhance an operation of more complicated slip patterns of dislocations (differences in operated slip systems and amount of slips), leading to a more inhomogeneous deformation. The inhomogeneous deformation may produce complicated crystal rotation place to place, and therefore introduce deformation-induced high-angle boundaries more frequently. Since solute atoms and precipitates can also suppress boundary migration or recovery of dislocations, ultrafine grains surrounded by deformation-induced high-angle boundaries can remain. These narrowly-spaced deformation-induced high-angle boundaries may also act as alternative sites for inhomogeneous deformation during further deformation, leading to a more complicated crystal rotation. Therefore, the obtained texture may become more randomized, and a higher fraction of high-angle boundaries could be introduced in the 2N-Al than in the 4N-Al.

5. Conclusions

In this study, 99.2% Al (2N-Al), 99.99% Al (4N-Al) and 99.999% Al (5N-Al) were deformed to high strains by ARB, and changes in microstructure and mechanical properties were investigated systematically. The main results obtained are as follows.

• (1)   In the 2N-Al and 4N-Al, 6 cycles of ARB process led to a quite uniform nanostructures dominated by high-angle boundaries throughout the sheet thickness. On the other hand, in the 5N-Al, nanostructures were not obtained due to the occurrence of recrystallization during the ARB. It was found that an increase in the amount of impurities was effective to increase the stability of the nanostructures and a purity of 99.99 mass% or less was required to obtain nanostructures by the SPD in pure aluminium.
• (2)   Deformation-induced grain subdivision process was enhanced by the redundant shear strain caused by the non-lubricated rolling. A good correlation was observed between the total equivalent strain and structural parameters of the obtained structures. This indicated that the total equivalent strain was a dominant parameter to determine the microstructural evolution during the ARB process.
• (3)   Deformation texture in the 2N-Al was more randomized than in the 4N-Al after 6-cycle ARB, resulting in a higher fraction of high-angle boundaries in the 2N-Al. Impurities existing as solute atoms or precipitates and deformation-induced high-angle boundaries were considered to enhance inhomogeneous deformation, responsible for an operation of more complicated slip patterns and resultant texture randomization.

Acknowledgements

This research was financially supported partly by the Grant-in-Aid for Scientific Research on Innovative Area “Bulk Nanostructured Metals” (contract No.22102006) through the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan, which is gratefully appreciated.

REFERENCES

• 1)  e.g. M.A. Meyers, A. Mishra and D.J. Bacon: Prog. Mater. Sci. 51 (2006) 427–556. 10.1016/j.pmatsci.2005.08.003
• 2)  D.A. Hughes and N. Hansen: Acta Mater. 45 (1997) 3871–3886. 10.1016/S1359-6454(97)00027-X
• 3)  N. Hansen and D. Juul Jensen: Philos. Trans. R. Soc. Lond. A 357 (1999) 1447–1469. 10.1098/rsta.1999.0384
• 4)  A. Gholinia, F.J. Humphreys and P.B. Prangnell: Acta Mater. 50 (2002) 4461–4476. 10.1016/S1359-6454(02)00253-7
• 5)  H. Jazaeri and F.J. Humphreys: Acta Mater. 52 (2004) 3239–3250. 10.1016/j.actamat.2004.03.030
• 6)  N. Tsuji, N. Kamikawa and B. Li: Mater. Sci. Forum 539–543 (2007) 2837–2842. 10.4028/www.scientific.net/MSF.539-543.2837
• 7)  A. Vorhauer and R. Pippan: Metall. Mater. Trans., A Phys. Metall. Mater. Sci. 39 (2008) 417–429. 10.1007/s11661-007-9413-1
• 8)  N. Kamikawa and N. Tsuji: Mater. Trans. 53 (2012) 30–37. 10.2320/matertrans.MD201115
• 9)  Y. Saito, H. Utsunomiya, N. Tsuji and T. Sakai: Acta Mater. 47 (1999) 579–583. 10.1016/S1359-6454(98)00365-6
• 10)  S.H. Lee, Y. Saito, N. Tsuji, H. Utsunomiya and T. Sakai: Scr. Mater. 46 (2002) 281–285. 10.1016/S1359-6462(01)01239-8
• 11)  N. Kamikawa, T. Sakai and N. Tsuji: Acta Mater. 55 (2007) 5873–5888. 10.1016/j.actamat.2007.07.002
• 12)  N. Kamikawa, N. Tsuji, X. Huang and N. Hansen: Mater. Trans. 48 (2007) 1978–1985. 10.2320/matertrans.MA200702
• 13)  R.K. Ham: Philos. Mag. 6 (1961) 1183–1184. 10.1080/14786436108239679
• 14)  X. Huang, N. Kamikawa and N. Hansen: Mater. Sci. Eng. A 483–484 (2008) 102–104. 10.1016/j.msea.2006.10.173
• 15)  N. Kamikawa, X. Huang, N. Tsuji and N. Hansen: Acta Mater. 57 (2009) 4198–4208. 10.1016/j.actamat.2009.05.017
• 16)  Y. Saito, N. Tsuji, H. Utsunomiya, T. Sakai and R.G. Hong: Scr. Mater. 39 (1998) 1221–1227. 10.1016/S1359-6462(98)00302-9
• 17)  Y. Ito and Z. Horita: Mater. Sci. Eng. A 503 (2009) 32–36. 10.1016/j.msea.2008.03.055
• 18)  Z. Horita, K. Kishikawa, K. Kimura, K. Tatsumi and T.G. Langdon: Mater. Sci. Forum 558–559 (2007) 1273–1278. 10.4028/www.scientific.net/MSF.558-559.1273
• 19)  N. Kamikawa and T. Furuhara: J. Mater. Process. Technol. 213 (2013) 1412–1418. 10.1016/j.jmatprotec.2013.02.016
• 20)  H.W. Höppel, J. May and M. Göken: Adv. Eng. Mater. 6 (2004) 781–784. 10.1002/adem.200306582
• 21)  H.-W. Kim, S.-B. Kang, N. Tsuji and Y. Minamino: Acta Mater. 53 (2005) 1737–1749. 10.1016/j.actamat.2004.12.022
• 22)  T. Sakai, Y. Saito, K. Hirano and K. Kato: Trans. ISIJ 28 (1988) 1028–1035. 10.2355/isijinternational1966.28.1028
• 23)  U.F. Kocks, C.N. Tomé and H.-R. Wenk: Texture and Anisotropy: Preferred Orientations in Polycrystals and their Effects on Materials Properties, (Cambridge University Press, UK, 1998), 178–239.
• 24)  W. Truszkowski, J. Król and B. Major: Metall. Trans. A 11 (1980) 749–758. 10.1007/BF02661204