Formation of Peripheral Coarse Grain in Thin-Walled Al–Mg–Si Extrusion Profiles

As sample material five experimental alloys of the Al–Mg–Si class have been designed that aim to have a medium to high strength. Their chemical composition is listed in Table I. With increasing amount of the hardening-phase forming elements Mg, Si and Cu, the weight fraction of dispersoid forming elements (DFE) Fe, Mn and Cr was increased as well. The aim was to achieve a monotonic increase of the amount of constituent particles and dispersoids. An exception to this is alloy 6005A.CP- which was designed to contain less constituent particles than 6005A by a reduced Fe content, while retaining a similar level of dispersoids.

All alloys have been cast by Direct Chill casting into 10” billets of 7 m length and were subsequently homogenized at 560 °C ± 20 °C for 4 hours by Hammerer Aluminium Industries Casting GmbH. After homogenization the billets have been extruded with a 41 MN servo-hydraulic extrusion press by Hammerer Aluminium Industries Extrusion GmbH. The billets were cut into blocks of 750 mm length and preheated to 490 °C before loading into the extrusion press. The blocks were then extruded into a thin-walled hollow profile with two chambers (Figure 1) with a wall thickness of 2.5 mm and a cross section area \({A}_{\rm{P}}\) of 1080 mm². In respect of the billet cross-section area \({A}_{\rm{B}}\) this results in a nominal extrusion ration \({A}_{\rm{B}}/{A}_{\rm{P}}\) of 47 and a corresponding deformation strain \(\varphi =\rm{ln}{A}_{\rm{B}}/{A}_{\rm{P}}\) of 3.9. The extrusions were conducted with three ram speeds to realize different strain rates during shape forming. This resulted in distinct profile exit speeds of 5, 10 and 15 m/min with profile exit temperatures of 528 °C ± 2 °C, 556 °C ± 5 °C and 570 °C ± 9 °C, respectively. The temperature was measured by a thermocouple in contact with the profile upon its exit of the press.

After exiting the extrusion press the profiles were water quenched with a minimum quench rate of > 20 K/s to retain the supersaturated solid solution. As the final processing steps, the profiles were stretched, cut and stabilization heat treated after extrusion.

For microstructural analysis of the dispersoids samples from profiles extruded at moderate extrusion speed (10 m/min) were taken. The dispersoids form during homogenization and are not altered by extrusion speed or temperature. Therefore, the microstructure obtained by the moderate extrusion speed is regarded as representative for each alloy in general. The sections were made perpendicular to the extrusion direction (EX), so that the transverse (TD) and normal direction (ND) lie within the analyzed plane. The sections were embedded in a polymeric resin, ground with SiC paper to 2400 grit and consecutively polished with diamond suspensions of 6, 3 and 1 µm particle size. The final polishing step was conducted with colloidal silica of 0.04 µm particle size.

Investigations and analysis of the dispersoid distribution of each alloy were conducted following the routine of Österreicher et al.^[25]: for each alloy, the microstructure was recorded in a total acquisition area of 1875 µm² in backscatter electron (BSE) contrast at 5 kV using a Zeiss 1540 EsB Crossbeam SEM. With ImageJ^[62] the acquired images were filtered, binarized and measured concerning the area and size of each dispersoid. In order to obtain information on the volume fraction, the registration depth of backscattered electrons was estimated using the CASINO Monte Carlo software.^[63] According to the routine proposed by Österreicher et al. explained in detail in^[25] a dispersoid size dependent registration volume was assessed for each particle. With the assumption of spherical dispersoid particles the volume fraction of dispersoids \({f}_{\rm{v}}\) was determined for each alloy following the summation in Eq. [2] over all dispersoids with a cut-off diameter \({d}_{\rm{cut}}\) < 400 nm:where \({v}_{i}\) is the volume of an individual dispersoid, \({V}_{i,\rm{acq}}\) is the corresponding acquisition volume, \({A}_{i}\) is the measured area of an individual dispersoid, \({z}_{i,\rm{reg}.}\) is its size dependent registration depth and \({A}_{\rm{acq}.}\) the total acquisition surface.

Constituent particles (CP) were quantified from sections of the extruded profile with orientations of 0, 25, 37, 45, 56 and 90 deg. The orientation angle describes the angle between the normal of the acquisition plane and the extrusion direction in the EX-TD-plane. The different orientation angles account for the preferred orientation parallel to the extrusion direction. Similar to the procedure used for dispersoid analysis, only the states processed at the moderate extrusion speed (10 m/min) were investigated. For each alloy and each orientation an acquisition surface of 0.75 mm² was recorded with backscatter electron (BSE) contrast at 12.5 kV using a Zeiss 1540 EsB Crossbeam SEM. The CP were evaluated with ImageJ^[62] to obtain the area fraction for each alloy and orientation. By incorporating particle analysis under different acquisition orientations, this results in a random particle distribution, where the area fraction can be regarded as the statistical mean of the volume fraction.

In order to compare the experimentally determined fractions of CP and dispersoids with their respective theoretically possible fractions at the homogenization temperature, equilibrium thermodynamic calculations were conducted using the aluminum database in JMatPro V10.2.

To obtain information on texture formation and grain size electron backscatter diffraction (EBSD) measurements were conducted in three planes per alloy and extrusion speed. The position of the selected specimens in the profile is drawn as a red box in Figure 1. The sections were chosen so that the normal of the investigated planes were parallel to EX, TD and the surface (ND). It is noteworthy, that directions TD and ND are defined in respect to the local specimen position where the sample was taken. Metallographic preparation consisted of grinding with SiC paper to 2400 grit, 6 µm diamond suspension polishing and final electrolyte polishing using an alcohol-based solution of 3 pct HClO₄ cooled to − 5 °C, polished for 5–10 seconds at 55 V. Sample preparation for the surfaces consisted solely of electrolyte polishing. For in depth analysis, measuring over a statistically representative number of grains, target grinding in ND direction was performed on selected samples.

The EBSD measurements were performed using a Zeiss 1540 EsB Crossbeam SEM with an Oxford Nordlys Nano detector. The beam step size was 3 µm for orientation analysis. The mappings in planes EX and TD had a width of 3.5 mm and a height of 2.5 mm, covering the whole wall thickness. The mappings in ND where sized to fit the measurements in the EX and TD planes. In case fine-grained microstructures could not be resolved sufficiently additional measurements limited to the fine-grained area were performed with as step size of 0.7 µm.

The EBSD measurements were used to analyze the micro- and macro-texture using the MTEX toolbox for Matlab.^[64] For the micro-texture analysis inverse pole figure (IPF) maps were created for each section plane EX, TD and ND. The reference orientation for color mapping was its plane normal EX, TD and ND, respectively. Additionally, low angle (2–10 deg) and high angle (>10 deg) grain boundaries were reconstructed for each mapping. The three perpendicular IPF maps were then arranged to a microstructure cube, to resemble a cubic cut as seen in Figure 1.

For the macro-texture analysis, the orientation distribution for each alloy and extrusion speed was quantified in terms of orientation fibers and texture components. The orientation fibers and texture components of interest are listed in Table II. The quantification was performed on the unfiltered mappings due to the high indexing rate of > 95 pct. For this, the specimen orientation was centered so that the orientation distribution coincided with the plane strain deformation mirror planes. This eliminates small deviations from ideal orientation introduced during grinding and mounting in the SEM, thus being necessary for texture quantification in terms of ideal texture components defined in Table II.

The orientations of each EBSD mapping were then compared to each of the ideal texture components and fibers by means of smallest misorientation angle. An ideal orientation or orientation fiber was assigned if the misorientation was \(\le 10 \, \rm{deg}\), which ensured that ideal texture components did not overlap. Consequently, each orientation were assigned to a single ideal texture component or were defined as a random orientation. Though, this does not generally apply to orientation fibers, e.g., the Goss texture component is both part of the \(\langle 001\rangle ||\rm{EX}\) and the \(\langle 101\rangle ||\rm{TD}\) orientation fiber. Summation over each texture component or fiber in relation to all indexed data points then yielded the fraction of each texture feature representing macro-texture of the measured cross section. Additionally, the spatial extent of the PCG layer was determined geometrically from the IPF maps in EX and TD section planes using ImageJ.

Finally, for texture and PCG layer representation of the profile microstructure, the means of the fractions of texture components from both EX and TD sections were taken.

The microstructures after extrusion of the alloys 6005 and 6082+ are shown in Figure 2(a) and (b), respectively. The phase distribution can be divided into large precipitates, constituent particles (CP), with a diameter of > 1 µm, and smaller dispersoids. The constituent particles are elliptical to elongated in shape. They form during casting and are broken down during extrusion. Dispersoids are of smaller size and precipitate during homogenization. Their precipitation is aimed to result in a diameter of < 400 nm, and they exhibit a spherical to elliptical shape. In Figure 2(a) the strong grain boundary pinning effect caused by the Smith–Zener drag of the dispersoids can be clearly seen. At a closer look at the grain boundary of the dark grain in the middle, pinning along its grain boundary on dispersoids and an inwards curvature between those points is visible. This is a sign for grain growth, that is consuming the dark grain, but is counter-acted by the Smith–Zener drag. Alloy 6005 in Figure 2(a) shows a smaller dispersoid density with 0.34 pct DFE than alloy 6082+ in Figure 2(b) with 1.13 pct DFE. This shows qualitatively that with an increasing content of DFE, a higher amount of nanoscopic dispersoids are formed.

The volume fraction of dispersoids in the investigated alloys is shown in Figure 3(a). For all five alloy compositions the dispersoid volume fraction increases following a logarithmic normal distribution towards a saturation at dispersoid diameters of about 400 nm. Although the volume fraction varies substantially between the alloys, the corresponding size distribution is relatively similar (Figure 3(b)). The cumulative frequencies of dispersoid size follows a logarithmic normal distribution with a median value for all alloys of 71–88 nm diameter. To avoid misconception, the distribution of cumulative volume fraction in Figure 3(a) accounts for the volume fraction of dispersoids and is therefore shifted to larger particle size, as small particles contribute less to overall volume fraction. By contrast, the cumulative relative frequency in Figure 3(b) regards numeric frequency, where each particle contributes equally, irrespective of its size.

The saturation at a dispersoid diameter of 400 nm in Figure 3(a) marks the dispersoid volume fraction of the alloys. The corresponding median dispersoid size for each alloy is defined by the dispersoid diameter at the median in Figure 3(b). These dispersoid values are listed in Table III together with the volume fraction of the constituent particles and the calculated thermodynamic equilibrium volume fraction. Figure 4 shows the increase in volume fraction of the equilibrium phase with increasing DFE content. The fraction of dispersoid and the CP fraction increases, accordingly. An exemption to this behavior is 6005A.CP-. It is compositionally close to alloy 6005A with a comparable content of DFE. However, due to a different casting routine and an adjusted composition of DFE this alloy shows less formation of CP during casting. This leads to an increased volume fraction of dispersoids by 18 pct to that of 6005A, as less DFE is incorporated into CP.

The offset between the sum of CP and dispersoids to the calculated equilibrium fraction in Figure 4 can be explained by the non-equilibrium state of the alloys. However, the finite resolution of SEM imaging could lead to an underestimation of volume fraction, as the smallest dispersoids < 20 nm could not be detected, although their contribution to overall volume fraction should be very small (Figure 3(a)).

The microstructure at the flat section of the profile as marked in Figure 1 is displayed in Figure 5 as microstructure cubes for the five investigated alloys, each extruded at three different extrusion speeds. Each microstructure cube consists of IPF maps of sections of the EX- and TD-plane and the profile’s surface as the ND-plane. Depending on extrusion speed and alloy different microstructural developments can be identified in the bulk of the cross section in the ODFs of Figure 6. Regarding the formation of microstructure in the bulk, alloy 6005 extruded at any extrusion speed, alloy 6005A extruded at 10 and 15 m/min and alloy 6005A.CP- at 15 m/min show many grains oriented along the \(\langle 001\rangle \) crystallographic axis lying parallel to the extrusion direction. On the other hand, alloy 6082+ at any extrusion speed, 6082 at 5 m/min and 10 m/min and 6005A.CP- at 5 m/min develop a dominant pole spread from \(\langle 211\rangle \) to \(\langle 111\rangle \) parallel to the extrusion direction, while retaining a small grain size. The remaining bulk microstructures in between can be described as a combination of both microstructure developments with small grains of \(\langle 211\rangle \)- to \(\langle 111\rangle \)-poles parallel to EX and large grains of \(\langle 001\rangle ||\rm{EX}\) orientation.

To further describe the texture development, additionally to the microstructure cube, Figure 6 shows the orientation density function (ODF) map for \({\varphi }_{2}\)-sections of 0, 45 and 65 deg of the whole cross-section of alloys 6005A (top row), 6005A.CP- (2nd row), 6082 (3rd row) and 6082+ (bottom row) extruded at 10 m/min. Figure 7 contains the texture development along the cross-section from surface to surface of the profile section in the sense of texture fiber content.

The ODF map of 6005A (Figure 6, top row) shows that its texture is dominated by the Cube texture component with a small spread of orientation to the Goss component. This is in line with the fiber description as \(\langle 001\rangle ||\rm{EX}\). This texture is associated with recrystallization texture of aluminum under plane strain and can be explained by the small amount of dispersoids in this alloy, thus exerting the smallest Smith–Zener drag on migrating grain boundaries enabling grain growth. However, global ODF does not sufficiently describe the local micro-texture and the through-thickness variation of orientation. Next to the bulk microstructure the peripheral coarse grain layer forms. This layer can be distinguished from the bulk texture in the TD plane of the microstructure cubes with \(\langle 101\rangle ||\rm{TD}\) oriented grains (Figure 7(a)). The distribution of texture fiber content along the cross-section consequently results in a dominant \(\langle 001\rangle ||\rm{EX}\)-fiber in the bulk while the PCG layer is defined by the \(\langle 101\rangle ||\rm{TD}\)-fiber. It must be noted, that at 400 µm depth both fibers add up to > 100 pct. This is due to the Goss texture component being part of both fibers.

The highly alloyed 6082+ (Figure 6, bottom row) globally displays a more homogenous distribution in the microstructure cube. The corresponding ODF shows a dominant \(\beta \)-fiber orientation with a strong peak at the Brass component and decreasing intensity over the S component to the Copper component, while the maximum overall intensity is smaller than that of 6005A. This is coherent with the observation from the microstructure cubes in Figure 5, which show the \(\langle 211\rangle \)- to \(\langle 111\rangle \)-poles parallel to EX. This \(\beta \)-fiber texture describes a deformation texture under plane strain in Al alloys.

The high amount of dispersoids in alloy 6082+ leads to a strong pinning of grain boundaries, effectively reducing grain boundary migration, resulting in a fine-grained microstructure with thin grain layers partially consisting of only one subgrain in height. A very thin PCG layer can be observed in the microstructure cube, but the intensity is too weak to be found in the ODF map. However, the cube component shows a weak but noticeable intensity in the ODF. This is reflected in the fiber orientation distribution for 6082+ (Figure 7(d)). The PCG layer only leads to a high fraction of \(\langle 101\rangle ||\rm{TD}\)-fiber just few tens of microns beneath the surface. The remaining cross-section is dominated by the \(\beta \)-fiber orientation in the bulk. The weak intensity related to the Cube component can be found at a depth of 750–1000 µm beneath the surface with a local maximum of the \(\langle 001\rangle ||\rm{EX}\)-fiber orientation. This local maximum correlates with a local minimum of the \(\beta \)-fiber, indicating that these bulk fiber orientations behave complementary. At a depth of 2300 µm a local maximum appears in the \(\langle 101\rangle ||\rm{TD}\)-fiber and adds up to > 100 pct together with the \(\beta \)-fiber. This is associated to the formation of the Copper component, which can be assigned to both fiber orientations.

Alloy 6082 extruded at 10 m/min (Figure 6, 3rd row) shows a higher number of local peaks in its global ODF map. The overall intensity is smaller than that of 6005A and 6082+, but similar to 6082+  the strongest intensity can be located at the Brass component with decreasing intensity to the S and the Copper component. However, additional peaks appear in the ODF map along the line of the Euler angles \({{\varphi }_{1}/\phi /\varphi }_{2}=90 \,{\rm{deg}}/x/45\, \rm{deg}\), describing the \(\langle 101\rangle ||\rm{TD}\)-fiber. The fiber orientation distribution along the cross section (Figure 7(c)) shows, that this fiber orientation is once more associated with the PCG layer. Its ODF reveals that, alongside with the Goss and Copper texture component, the strongest intensities of the \(\langle 101\rangle ||\rm{TD}\)-fiber are found at the Rotated Cube and the Copper Twin texture component. In between the PCG layers the bulk is defined by a sharp decrease of the \(\langle 101\rangle ||\rm{TD}\)-fiber content and a dominant \(\beta \)-fiber fraction, comparable to 6082+. In contrast to 6082+, the fraction of \(\langle 001\rangle ||\rm{EX}\)-fiber increases for 6082. Due to this complementary behavior of the fiber’s texture components the maximum \(\beta \)-fiber fraction is reduced.

The microstructure cube of 6005A.CP- (Figure 6, 2nd row) shows a large PCG layer comparable to 6082, as discussed above. The bulk shows large grains of \(\langle 001\rangle ||\rm{EX}\)-fiber orientation separated by a small fraction of small \(\beta \)-fiber oriented grains. According to this, the ODF shows a large extent of \(\langle 001\rangle ||\rm{EX}\)-fiber with a strong Cube texture component, a minor peak at the Brass component and a split of \(\langle 101\rangle ||\rm{TD}\)-fiber into Rotated Cube and Copper Twin component.

This separation of grain sizes in the bulk with their corresponding fiber orientations describes the transition microstructure between none and extensive grain growth of \(\langle 001\rangle ||\rm{EX}\) oriented grains. While in the bulk two kinds of microstructures can develop, the PCG layer is described by the development of the \(\langle 101\rangle ||\rm{TD}\)-fiber in the global ODF. The through-thickness transition from the PCG layer to the bulk is visible as a less steep slope in the orientation fiber fraction distribution (Figure 7(b)) at 500 − 600 µm due to the formation of the Goss texture component, that is part of the \(\langle 001\rangle ||\rm{EX}\) as well. Additionally, the grains of the PCG layer and the \(\langle 001\rangle ||\rm{EX}\) oriented grains in the bulk are of comparable size. This leads to a less even and planar boundary between PCG layer and bulk.

Therefore, the bulk and PCG layer can be distinguished by the fiber components. In addition to that, the bulk can be characterized regarding the two fiber orientations \(\langle 001\rangle ||\rm{EX}\) and \(\beta \). Figure 8 shows the fraction of microstructure components over the whole cross section of each alloy extruded with 5 m/min (left), 10 m/min (middle) and 15 m/min (right). With increasing extrusion speed the deformation rate increases, leading to higher strain rates and higher temperatures of the extrusion profile. This increases the grain boundary mobility that enables grain growth. The strain distribution over the cross section, that affects the stored energy and thus the driving force for grain growth, however, is expected to stay relatively constant for the different extrusion speeds. In Figure 8 the alloys are arranged in the order of increasing dispersoid content and hence grain boundary pinning effect, which counteracts the grain boundary mobility. The amount of the PCG layer is characterized by its depth over the cross section and texture information. The microstructure composition in the bulk is shown as \(\langle 001\rangle ||\rm{EX}\)- and \(\beta \)-fiber fraction, describing the coarse-grained and fine-grained microstructure, respectively. For all investigated microstructures there is only a very small amount (< 15 pct) of random orientation in the bulk that is not part of any of the two orientation fibers. This emphasizes the strongly textured nature of the investigated alloys upon extrusion shape forming.

At the low extrusion speed of 5 m/min and hence small deformation rates all alloys, except for 6005, exhibit \(\beta \)-fiber in the bulk (Figure 8, 5 m/min). The amount of \(\langle 001\rangle ||\rm{EX}\)-fiber stays constant at an overall fraction between 10 and 15 pct. While in alloy 6082+, with the highest dispersoid content, nearly no PCG forms, the extent of PCG increases with reducing dispersoid content. The largest PCG size at this extrusion speed is reached in 6005A. Further reduction of the amount of dispersoids in 6005 does not increase the size of the PCG layer but increases the fraction of \(\langle 001\rangle ||\rm{EX}\)-fiber in the bulk at the expense of the \(\beta \)-fiber, which is not present in this microstructure.

With an increase of the extrusion speed to 10 m/min (Figure 8, 10 m/min) or to 15 m/min (Figure 8, 15 m/min), the largest PCG is observed in 6005A.CP- and 6082, respectively. At these speeds the bulk of 6005A and 6005A.CP- changes to a coarse-grained \(\langle 001\rangle ||\rm{EX}\)-fiber microstructure. In 6082 and 6082+ the fraction of fine-grained \(\beta \)-fiber-oriented microstructure in the bulk is reduced and the extent of PCG grows. It is noteworthy that when the bulk microstructure is mainly occupied by the \(\beta \)-fiber microstructure, the fraction of \(\langle 001\rangle ||\rm{EX}\)-fiber in the bulk stays constant at a level of 10−15 pct.

To further demonstrate the effect of the dispersoid content, deformation rate and temperature on the formation of microstructural components, Figure 9 shows the evolution of microstructure with increasing extrusion speed for the three alloys 6005A, 6005.CP- and 6082. The increase in extrusion speed increases the profile temperature, which leads to a higher grain boundary mobility. In alloy 6082, with a dispersoid volume fraction of 0.55 pct, the increasing grain boundary mobility affects the growth of the PCG layer to the expense of the \(\beta \)-fiber in the bulk.

In alloy 6005A the grain boundary mobility is already high enough for the formation of a large PCG layer, because the smaller dispersoid volume fraction of 0.33 pct results in a less effective grain boundary pinning. However, the pinning in the bulk is large enough to suppress the growth of a coarse-grained \(\langle 001\rangle ||\rm{EX}\)-fiber microstructure. Only with higher extrusion speeds the grain boundary mobility is increased so that a coarse-grained \(\langle 001\rangle ||\rm{EX}\)-fiber microstructure in the bulk can form. Additionally, this has the effect that the PCG layer is of smaller dimension, reducing from a fraction of 0.53 at 5 m/min to 0.31 at 10 m/min and 0.24 at 15 m/min. Alloy 6005A.CP- has a volume fraction of dispersoids of 0.39 pct, which is between alloys 6005A and 6082. Moreover, it shows a smaller amount of CP, decreasing the number of sites for particle stimulated nucleation. Consequently, the fraction of microstructural components at 5 m/min lies between 6005A and 6082. Comparable to 6082 the PCG grows larger at 10 m/min extrusion speed, due to the increased grain boundary mobility. Additionally, growth of coarse \(\langle 001\rangle ||\rm{EX}\) oriented grains already started in the bulk, leaving only a small fraction of \(\beta \)-fiber microstructure in the bulk. At the highest extrusion speed of 15 m/min the fraction of \(\langle 001\rangle ||\rm{EX}\) microstructure increases further while the amount of PCG decreases.

The microstructural investigation shows that with increasing extrusion speed and reducing dispersoid content at first the PCG grows, consuming a large amount of the \(\beta \)-fiber bulk. The formation of the \(\langle 001\rangle ||\rm{EX}\)-fiber in the bulk then hinders further growth of the PCG layer into the bulk and counteracts its extensive growth. The formation of microstructure is dependent on the grain boundary mobility affected by the extrusion speed and temperature and the pinning pressure exerted by the dispersoids due to the Smith–Zener drag. However, the PCG growth cannot be regarded independently of the microstructural evolution in the bulk.

From Figure 8 it is obvious, that alloy 6082+ shows the highest resistance against grain growth, both in the bulk and the PCG layer. The microstructure of the cross-section extruded with a speed of 5 m/min shows few to no formation of a PCG layer. With increasing extrusion speed, however, a PCG layer grows sub-surface at the expense of the \(\beta \)-fiber in the bulk.

Figure 10(a) shows the sub-surface layer of 6082+  extruded at 5 m/min in the TD-ND plane, the orientations are colored according to the poles parallel to EX. Few larger grains form at the immediate surface without a specific texture. Going further into the material, a layer of small grains forms at a depth of 10 to 50 µm (Figure 10(c)), that exhibit an orientation close to \({{\varphi }_{1}/\phi /\varphi }_{2}=51\,\rm{ deg}/66\, \rm{deg}/63\, \rm{deg}\) or \(\left\{211\right\}\langle 101\rangle \), respectively. This orientation is also referred to as the B shear component.^[65] With increasing depth, the orientation of the grains gradually transitions to the \(\beta \)-fiber orientation of Brass and S texture components. The misorientation relation between the B shear component and the Brass and the S texture component, respectively, is 30 deg for both, but the rotational axis is \(\left[111\right]\) for Brass, which lies parallel to TD, and \(\left[3\,2\,12\right]\) for the S component, which is oriented diagonally in the EX-TD-ND coordinates. The gradual transition from the B shear component to the \(\beta \)-fiber by gradual rotation around their rotational axes is reflected in Figure 10(c). Texture components are defined as orientations within a misorientation of 10 deg. As B shear is about 30 deg misoriented to the \(\beta \)-fiber orientation, an interval of 10 deg in orientation space is not regarded between B shear and the \(\beta \)-fiber, so that the component fraction in Figure 10(c) cannot add up to 100 pct.

The microstructure of 6082+ extruded at the increased extrusion speed of 10 m/min shows more PCG formation in Figure 8, albeit with 3.5 pct to a small extent overall. The sub-surface microstructure shows the formation of large, recrystallized grains at the immediate surface (Figure 10(b)). From the sample’s orientation profile (Figure 10(d)) it is evident, that no B shear component forms. The grains of the PCG layer belong to the \(\langle 101\rangle ||\rm{TD}\) fiber. There is no gradual transition to the fine microstructure, but the \(\beta \)-fiber-oriented grains border abruptly to the PCG layer with increasing distance to the surface. Because the boundary between PCG layer and fine-grained microstructure is not parallel to the surface but uneven, the transition from PCG to fine-grained microstructure appears gradually in the texture profile in Figure 10(d). This boundary area lies 50 ± 20 µm beneath the surface. Figure 11(a) shows the microstructure of this boundary region in the EX-TD plane in 50 µm depth to the surface. The IPF map shows large, recrystallized grains of the PCG with a \(\langle 101\rangle ||\rm{TD}\) fiber orientation and fine grains with an elongated shape in EX with \(\langle 111\rangle \)-poles aligned parallel to TD. The ODF of this IPF map (Figure 11(b)) reveals that the fine-grained microstructure exhibits Brass component and the preferential orientation of PCG grains can be related to a composition of Rotated Cube, Goss and Copper Twin orientations. Between the latter two the ODF exhibits a spread rather than discrete peaks.

The alloy composition and extrusion speed influence the formation of distinct microstructures and textures. From Table III it becomes clear, that with an increase in DFE the volume fraction of dispersoids increases, while the median radius stays constant. This leads to a monotonic increase of the Smith–Zener drag pressure with increasing amount of DFE. This behavior is shown in Figure 12(a). As a consequence, this enables a more effective pinning of grain boundaries during extrusion. With increasing extrusion speed the global strain rate subjected to the material increases analogically, which in consequence rises the material temperature due to adiabatic heating. However, the local strain states can vary strongly both over the profile cross section as well as across the wall thickness. These strain states can only be determined numerically by material flow simulations. However, besides the necessary software, this requires a digital model of the extrusion tool geometry as well as a parameterized material model for each alloy to calculate the flow stresses and the resulting friction, which were not available for the present investigation. Therefore, in order to ensure comparable strain states the identical position on the profile was chosen for the microstructure analysis.

The increase in strain rate leads to an increase of profile temperature after exiting the extrusion die orifice. This eases thermally activated processes such as recovery of dislocations through climbing as well as grain boundary mobility. Dislocations accumulate into grain boundaries as described by CDRX and GDRX, which results in a fine-grained microstructure that determines the stored energy in the form of the grain boundary density. This in turn is associated with the strain distribution over the profile cross section and is expected to be relatively constant and independent of the extrusion speed. The thermally increased grain boundary mobility then enables grain growth following the driving force to reduce the stored energy. The balance between the driving forces for grain growth and the counter-acting pinning pressure result in characteristic microstructures represented by microstructure cubes displayed in Figure (5).

The highest pinning pressure is exhibited by 6082+ which also shows the highest dispersoid content. Therefore, the microstructure in the bulk consists of elongated grains of \(\beta \)-fiber orientation at all extrusion speeds (Figures 10(a) and (b)). The formation of such a fine-grained microstructure can be explained by the mechanisms of CDRX or GDRX. Although the differentiation between both is difficult, GDRX appears to be most likely. This is concluded because EBSD-measurements show that most grains exhibit a thickness of only 1-2 subgrains and the characteristic deformation texture is preserved. Additionally, without further regard of any superimposed shear strains, the nominal deformation strain \(\varphi \) of 3.9 alone exceeds the critical strain of 2.5 at which GDRX has been reported to occur.^[47,48]

Due to the relationship between grain structure and texture, the microstructure can be quantified by means of fraction of \(\beta \)-fiber orientation. In this fiber, orientations concentrate at the Brass component and show a spread over the S component, but only little of the Copper orientation component can be detected. Additionally, the formation of 10–15 pct Cube and \(\langle 001\rangle ||\rm{EX}\) oriented grains is consistently found together with the \(\beta \)-fiber orientation. The pronounced Brass formation as well as the small amount of Cube orientation is explained by activation of additional slip systems on the \(\left\{110\right\}\) and \(\left\{100\right\}\) planes during plane strain deformation at elevated homologous temperatures.^[66‐68]

Figure 12(b) shows the dependence of the \(\beta \)-fiber content in the bulk in dependence of the Smith–Zener relation for the extrusion speeds 5, 10 and 15 m/min. This relation is a measure of the grain boundary pinning effectiveness of dispersoids in the alloy system independent of the exact grain boundary energy (c.f. Eq. [1]), which depends on the misorientation between two grains. It is evident that a higher grain boundary pinning effect leads to a larger fraction of the deformation texture. However, the retained level of deformation texture is reduced by the increasing extrusion speed, due to a higher profile exit temperature, which promotes grain growth through enhanced grain boundary mobility. This enhanced grain boundary mobility affects the bulk in such a way that Cube oriented grains in Cube bands, formed during extrusion by the above explained mechanism, show preferential growth into the deformed microstructure due to high crystallographic symmetry. Besides less stored energy in Cube grains, preferential growth is supported by a near 40 deg misorientations around \(\langle 111\rangle \) of the Cube orientation to a large fraction of orientations of the \(\beta \)-fiber.^{[28,51,54‐56,69]} This applies to misorientations close to Σ7 grain boundaries, which pose a high mobility under growth conditions, although a slight deviation from the ideal 38.2 deg tilt misorientation around \(\langle 111\rangle \) of the highest mobility.^[70] The fraction of \(\langle 001\rangle ||\rm{EX}\) orientated grains, which incorporate Cube and Goss orientation components, as well as the spread in between, is shown in Figure 12(c) in dependence of the Smith–Zener relation. Here, an inverse relation between the fiber content and the pinning effect on grain boundaries can be observed. The fraction of \(\langle 001\rangle ||\rm{EX}\) orientated grains increases with higher thermal activation of grain boundary mobility, due to the higher extrusion speed. However, the behavior of the contents of \(\langle 001\rangle ||\rm{EX}\) and the \(\beta \)-fiber are not complementary, as they do not add up to 100 pct. This is due to the formation of the PCG layer, which to a large extent forms at intermediate pinning pressures (Figure 12(d)). With increasing extrusion speed, the peak of PCG layer fraction is shifted towards a higher pinning effectiveness. That means, that PCG formation can be suppressed by shifting the balance of driving and counter-acting forces to either extensive grain growth of Cube grains in the bulk or to high grain boundary pinning effectiveness due to dispersoids.

From the microstructure cubes (Figure 5) and their components (Figures 8 and 9) it becomes obvious, that with high pinning effects in 6082 and 6082+ the PCG layer grows deeper into the \(\beta \)-fiber microstructure with increasing grain boundary mobility, due to thermal activation. The content of \(\langle 001\rangle ||\rm{EX}\) oriented grains in the bulk remains constant, only for 6082 at 15 m/min the formation of larger grains of such orientation can be observed. This indicates that the grains in the PCG layer have a higher grain boundary mobility than that of Cube and other \(\langle 001\rangle ||\rm{EX}\) oriented grains in the bulk. This can be explained by friction between the material and the extrusion tool during material flow. This friction leads to a shear strain in the vicinity of the surface, which adds to the global plane strain deformation, that increases the adiabatic heating and thus the thermal activation of grain boundary mobility further compared to the bulk. The presence of shear strain underneath the surface also affects the microstructure evolution in alloys 6005A and 6005A.CP- at 5 m/min and 10 m/min. While the microstructures of both alloys at 5 m/min exhibit the dominating \(\beta \)-fiber microstructure in the bulk, a PCG layer forms beneath the surface, which is smaller for 6005A.CP- with a comparable Smith–Zener relation but smaller amount of CP, which promote recrystallization. Increasing the extrusion speed to 10 m/min rises grain boundary mobility in the bulk so that grains of Cube and \(\langle 001\rangle ||\rm{EX}\) orientations can grow into the surrounding \(\beta \)-fiber grains. In 6005A this results in the \(\langle 001\rangle ||\rm{EX}\) microstructure taking up the whole bulk and reducing the PCG extent, compared to the state at 5 m/min. In 6005A.CP- \(\langle 001\rangle ||\rm{EX}\) oriented grains are dominating in the bulk, but \(\beta \)-fiber microstructure is not completely consumed yet. Therefore, for this alloy the PCG layer has the largest dimension in this state. Further increase in extrusion speed to 15 m/min and thus temperature increase the fraction of \(\langle 001\rangle ||\rm{EX}\) grains and further reduces the PCG layer dimension.

With the growth of the PCG layer, a selection of specific orientations is evident, together with an increase in grain size from the surface towards the bulk. Hence, the mechanism of grain growth in the PCG layer appears to be of orientation selection of specific orientations during grain growth into the \(\beta \)-fiber textured bulk of grains with either Rotated (\(45 \,\rm{deg}||\rm{ND}\)) Cube, Goss or Copper Twin orientation, that originate in the surface or sub-surface layer. These orientations share the common crystallographic axis orientation \(\langle 101\rangle ||\rm{TD}\).

The Rotated Cube component has been reported as a characteristic shear deformation texture component apparent in aluminum, together with the B shear orientation.^[65] However, this shear deformation texture relates to lattice rotation upon deformation. Such shear deformation at elevated temperature is expected to form small grains by means of the GDRX mechanism, which is applicable only to the B shear orientation in the investigated alloys. Goss and Copper Twin components have so far not been associated with shear deformation texture. The increasing grain size in the PCG layer towards the bulk appear to follow a mechanism of orientation selection of those orientations. This can be explained by the nucleation of grains at the immediate surface, which is only weakly textured. After the profile exits the die orifice, the grains of Rotated Cube, Goss or Copper Twin orientation grow perpendicular to the surface through the shear deformed zone into the \(\beta \)-fiber textured microstructure. The parallel alignment of the boundary between the bulk and the PCG layer emphasizes the orientational relation of the PCG layer growth. While the driving force for the coarse grain growth can be clearly identified as the reduction in grain boundary energy of the fine-grained \(\beta \)-fiber microstructure, the reason for the selection of orientations of \(\langle 101\rangle ||\rm{TD}\) relation is not clear, as extensive growth of such oriented grains have not yet been reported. A special misorientation relationship between the texture components of the PCG layer to the orientations of the \(\beta \)-fiber as well as to the B shear component, respectively, are highly probable. Testing the misorientation between the three texture components Brass, S and Copper of the \(\beta \)-fiber against the three texture components of the PCG layer in terms of a rotation around the \(\langle 111\rangle \) axis reveals that the S component shows the smallest deviation angle of < 5 deg to the Copper Twin component at \(41 \,\rm{deg}\langle 111\rangle \). The misorientation of the Goss and Rotated Cube component to the S component show larger remaining deviation angle of about 14 and 18 deg, respectively, however, these mark the minimum deviation angles to the S component at \(\sim 40\, \rm{deg}\langle 111\rangle \). The deviation angle to the Σ7 misorientation between the remaining Brass and Copper texture component to the orientations of the PCG layer are much larger. Therefore, preferential growth in terms of Σ7 misoriented grain boundaries can only be probable for the growth of the Copper Twin component consuming the S component. Nonetheless, this is the identical relation as in one of the explanations for the dominant Cube recrystallization texture via highly mobile grain boundaries that promote grain growth. However, these are only the second-most prominent texture components in their respective microstructural zone. Most grains in the PCG layer are of Rotated Cube orientation, although no favorable misorientation to the dominant Brass component in the bulk can be distinguished. A low-energy Σ3 \(60 \,\rm{deg}\langle 111\rangle \) twinning relation is existent between Rotated Cube and Copper Twin that might stabilize the selection of both grains, although the high-stacking fault energy in aluminum demotes twinning. The orientation relations found so far can be regarded as first indications on to why favorable growth of grains occurs with distinct \(\langle 101\rangle ||\rm{TD}\) orientations, however, the prevailing mechanism can only be speculated about so far. Another influence on the kind of PCG formation is probable due to construction of the die orifice. In this study, an identical die was used for all alloys, to keep the deformation conditions and strain state constant. Mahmoodkani et al.^[39,40] have shown that adjustment to the die orifice design can reduce the formation of PCG due to smaller strain gradients beneath the surface, however, the shear deformation zone extends deeper into the material.