2019 Volume 60 Issue 2 Pages 322-329
The effect of cooling rate on microstructure of vertical-type high-speed twin-roll cast 3003 aluminum alloy strip was investigated. The solidification structure was characterized in terms of grain size and kinds, size, morphology and chemical composition of secondary particles. The 3003 aluminum alloy strip consisted of cell structure, dendritic structure, globular grains and eutectic structure along the strip thickness direction. From the relationship between the cooling rates and as-cast grain size, the cooling rate of high-speed twin-roll cast strip surface area was estimated as 3.1 × 103 K/s. Significant differences in the formation of secondary particles were found between direct chill (DC)-casting and the high-speed twin-roll casting as a result of the different cooling rates; Al6(Mn,Fe) and α-Al(Mn,Fe)Si phase were identified in the DC-cast sample, whereas only α-Al(Mn,Fe)Si was predominant in the high-speed twin-roll cast strip. The Al6(Mn,Fe) particles in DC-cast sample were script-like morphology with high aspect-ratio. In contrast, α-Al(Mn,Fe)Si particles in the high-speed twin-roll cast strip was spore-like morphology. Most α-Al(Mn,Fe)Si particles in the strip were considered to be formed as one of the eutectic components from the liquid droplets trapped in the inter-dendrite regions. In particular, Fe-rich α-Al(Mn,Fe)Si phase was formed at strip central area due to the increase in Fe segregation at the growth front.
Wrought Al–Mn based 3xxx alloys are mainly used for fin of automotive heat exchangers and thin sheets for beverage can body, because of their moderate strength, good ductility and corrosion resistance, etc. For manufacturing process, conventional direct chill (DC)-casting followed by homogenization, hot- and cold rolling has been applied. The DC-casting is commonly conducted in a range of cooling rates, less than 20 K/s.1) Due to the relatively low cooling rate of the DC-casting process, it is inevitable not only low Mn solubility that unfavorable to the formation of fine dispersoids by given sequence heat treatment, but also forming Mn-containing coarse precipitates in as-cast condition. The coarse secondary particles can act as initiation sites for cracking that lead to a deterioration of mechanical properties in the final rolled sheet or foil. Although the downstream processes are provided for the wrought aluminum alloys, the initial solidification structure of as-cast products affects the microstructure and physical and mechanical properties.
Recently, vertical-type high-speed twin-roll casting (HSTRC) has been paid attention as a promising manufacturing process to produce a thin aluminum alloy strip directly from the melt with a remarkable high casting speed.2–4) Compared with the conventional DC-casting, the HSTRC has numerous advantages; reduction in processing steps, equipment cost, operating cost and energy consumption. The HSTRC is characterized by high cooling rates.5) This feature is expected to lead to not only fine solidification structure including homogeneous distribution of fine secondary particles, but also enhanced solid solubility of solute atoms in the matrix. For dealing with increasing demand for thinner and stronger aluminum alloy sheets, the HSTRC is considered to be applicable to provide an ideal initial microstructure, which is a key to improve the final mechanical properties of Al–Mn based 3xxx alloy sheets.
In this study, commercial 3003 aluminum alloy strip was fabricated by HSTRC. The present work focused on the solidified structure and secondary particles in the as-cast condition, and conducted detailed metallographic observation. Through comparison of the microstructure between conventional DC-cast sample and HSTRC strip, the effect of high cooling rate of HSTRC on solidification structure including secondary particles was discussed.
The commercial 3003 aluminum alloy was used in the present study. The DC-cast billets of 3003 aluminum alloy were provided from the UACJ corporation. The chemical composition of 3003 aluminum alloy was (in mass%) Mn 1.26, Fe 0.61, Si 0.27 and Cu 0.15.
Figure 1 shows a schematic illustration of the vertical-type high-speed twin-roll caster. The caster was equipped with a pair of copper rolls. A diameter and width of the rolls were 300 mm and 100 mm, respectively. The rolls were cooled by running water. One of the rolls was firmly fixed to the base, while the other roll was installed with a series of springs to control the initial roll separating force. The roll rotation speed was 60 m/min. The initial roll separating force was set to be 20 kN. The minimum roll gap was 1 mm.
Schematic illustration of the vertical-type high-speed twin-roll caster.
The DC-cast billet was melted in the electric furnace, and provided for HSTRC. Before the casting, the melt was degassed by injecting Ar gas into the melt. The casting temperature was selected around 15 K over the liquidus temperature. Pouring the melt into the feeding nozzle, solidification shells were formed on the both roll surfaces. A feeding nozzle which was set over the rolls contributed to maintain the stable melt-height. The stable melt-height was necessary for the stable solidification shell formation on the both roll surfaces. Following the roll rotation, the solidified shells encountered at the roll gap, and were rapidly cooled. At last, a several meters-long thin strip was produced within a few seconds. In the present study, approximately 3–4 m-long and 100 mm-wide strip was fabricated from 2.5 kg molten alloy. The constant thickness, which was about 2.4 mm, was obtained in the middle part of the cast strip along the casting direction (around 1 to 2 m-long). It should be noted that the microstructure observation was performed on this part.
In order to examine the effect of cooling rates during solidification, the DC-cast billet was melted, and solidified under some different cooling rate conditions using a permanent steel mold (1∼10 K/s) and a copper mold (10∼100 K/s). The steel mold was designed to have a wedge-shape cavity to get different cooling rates depending on the local thickness of the cast plate.6) The cavity of copper mold was 100 mm × 100 mm × 2 mm in size. The cooling rate was obtained by the slope of cooling curve at the liquidus temperature.
The metallographic examination was conducted using OM, FE-SEM and TEM. The number density and area fraction of secondary particles were measured using Image-J software. The measurement was conducted from 10 images for FE-SEM (JEM-7000F, JEOL) observation. Each image was taken randomly from the as-polished sample. In order to observe 3-dimensional morphology of secondary particles, the samples were deep-etched by 5% NaOH solution. The detailed features were characterized by FE-SEM (JEM-7000F, JEOL). Phase identification of the secondary particles was performed using X-ray diffractometer (RINT-2100, Rigaku) with Cu Kα (λ = 1.54 Å) radiation. Chemical composition analysis of the secondary particles was also performed by using EDS attached to the TEM (JEM-3010, JEOL).
The cooling rates during solidification are often estimated from the solidified microstructure. Suzuki et al.7) estimated the cooling rate of HSTRC A356 alloy strip using dendrite arm spacing (DAS) of primary α-Al dendrite. Shimosaka et al.8) estimated the cooling rate of HSTRC Al–Mg–Si–Mn alloy strip by DAS and inter lamellar spacing (ILS) of its eutectic structure. In contrast to that, it is difficult to observe the clear DAS or ILS in solidified structure of Al–Mn based alloys including 3003 aluminum alloy. Therefore, instead of using DAS or ILS, the relationship between the cooling rate and grain size was examined. Figure 2(a) to (f) shows the polarized light micrographs of permanent mold cast sample, DC-cast sample and HSTRC strip. It can be seen that the grain size reduced as increasing the cooling rate. The DC-cast sample showed around 90–130 µm in grain size, whereas near-surface area of the HSTRC strip showed around 15–19 µm in grain size. In the case of HSTRC strip, the grain size was different along its thickness direction. The near-surface area is considered having higher cooling rate rather than that of the strip central area. Additionally, solidified microstructure of the central area is a mixture of equiaxed and globular grains. It is hard to evaluate the grain size in this area. Thus, in the present study, only the near-surface area was provided for the examination of the effect of cooling rate on the grain size. The relationship between grain size and cooling rate is shown in Fig. 2(g). The grain size was well approximated on a straight line with a slope of −0.327 as a logarithmic scale. By using the present result, the cooling rate of the DC-casting and HSTRC was estimated. The estimated cooling rate was around 11 K/s for DC-casting, 3.1 × 103 K/s for HSTRC strip at near-surface area, respectively. This result showed good agreement with the directly measured cooling rates by using the thermocouples in the previous work.5)
Polarized light micrographs of the wedge-shape steel mold cast sample (a) top, (b) middle, (c) tip area, (d) DC-cast sample, (e) copper mold cast sample, (f) HSTRC strip near-surface; (g) effect of cooling rate on as-cast grain size for 3003 aluminum alloy.
Figure 3 shows micrographs of the as-cast 3003 aluminum alloy strip. Columnar grains grew from both strip surfaces, and they were gradually replaced by the equiaxed grains along the thickness direction as shown in Fig. 3(a). In the mid-thickness region, a band of fine globular grains, so-called a central band, was observed. This is a common microstructural feature of the HSTRC strips of the alloys, which exhibit mushy-type solidification.4) In the thickness direction, the strip had a clearly different microstructure (Fig. 3(b)). Near the strip surface region, cell structure was observed. Inside of the each cell, fine etched pits were observed as shown in Fig. 3(c). They correspond to the locations of homogeneously dispersed secondary particles. The cell structure gradually changed into the clear dendritic structure along the strip thickness direction as shown in Fig. 3(d). In the mid-thickness area, globular grains and eutectic structure were observed as shown in Fig. 3(e).
Optical micrographs of as-cast 3003 aluminum alloy strip. (a) Grain structure of the transverse cross section of the strip, (b) solidified structure of half thickness area, (c) near-surface area, (d) dendrite area, (e) mid-thickness area.
Figure 4 shows micrographs of the DC-cast sample and near-surface of the HSTRC strip. In DC-cast sample (Fig. 4(a)), secondary particles were mainly located at inter-dendrites. Most of the secondary particles in DC-cast sample were script-like in 2-dimensional morphology with high aspect ratio (Fig. 4(c)). In case of the HSTRC strip, both α-Al cells and secondary particles were much finer than those of DC-cast sample (Fig. 4(b)). The secondary particles were located inside of the cells as well as grain boundaries as shown in Fig. 4(d). Compared with the DC-cast sample, the secondary particles in HSTRC strip exhibited spherical-shape inside of the cells and elongated shape along the boundaries, respectively.
Secondary particles distribution in DC-cast sample and HSTRC strip. OM images; (a) DC-cast sample, (b) HSTRC strip near-surface. Back-scattered electron images; (c) DC-cast sample, (d) HSTRC strip near-surface.
Significant refinement of secondary particles was detected in HSTRC strip. However, due to the distinct shape differences in secondary particles, (i.e., particles between DC-casting and HSTRC, and between grain interiors and boundaries), it was difficult to compare the particle size distribution directly. Figure 5 shows the number density and area fraction of the secondary particles in DC-cast sample and HSTRC strip surface. It should be noted that due to significant difference in grain size as well as particle size and its distribution, different magnification was applied to DC-cast sample and HSTRC strip. As a result, the observed area per one image was different, 10718 µm2 and 4266 µm2, respectively. In the particle area fraction, the difference between DC-cast sample and HSTRC strip surface was relatively small, while the difference in particle number density was large between DC-cast sample and HSTRC strip surface. Since the grain size of the HSTRC strip is quite small (see Fig. 2), the relative amounts of grain boundaries providing the formation site of secondary particles are large. In addition, due to the high cooling rate of HSTRC, the residual liquid droplets are trapped at the inter-dendritic regions. This is considered to result in the homogeneous distribution of refined secondary particles in as-cast HSTRC strip.
Comparison of number density and area fraction of the secondary particles in DC-cast sample and HSTRC strip surface.
Figure 6 shows the XRD patterns for the DC-cast sample and HSTRC strip of as-cast condition. As expected from the Al–Mn–Fe–Si phase diagram,9) Al6(Mn,Fe) phase with Al matrix phase was detected in the DC-cast sample. Small peaks of α-Al(Mn,Fe)Si phase were also detected. In contrast to that, the α-Al(Mn,Fe)Si phase was dominant in HSTRC strip on both strip surface and central area.
XRD patterns of 3003 aluminum alloy. (a) HSTRC strip near-surface, (b) HSTRC strip central area, (c) DC-cast sample.
Solute elements (Mn, Fe and Si in 3003 aluminum alloy) are considered to segregate at the solid/liquid interface during solidification. Compositions of the solute-rich liquid can be varied depending on the solid solubility of each solute element in Al matrix and the solidification rate. Difference in composition of the liquid determines the sort of secondary-phase. In 3xxx (based on Al–Mn–Fe–Si alloys) alloys, Fe element favors the precipitation of Al6(Mn,Fe) phase, whereas Si element results in the formation of α-Al(Mn,Fe)Si phase. Alexander and Greer10) reported the compositional change in the liquid during solidification in Al–0.5Fe–1.0Mn–0.2Si (in mass%) alloy using MTDATA thermodynamic modeling software under Scheil approximation (i.e., assuming a uniform solute distribution in the liquid and no diffusion occurs in the solid). As the large undercooling, solute partitioning pushes Fe and Si into the liquid. The undercooling increases with high cooling rate during solidification. The HSTRC has much higher cooling rate compared with that of DC-casting. It can lead to substantial partitioning of solute elements including Si into the liquid. The enrichment of Si in liquid probably results in the formation of the α-Al(Mn,Fe)Si phase in HSTRC strip rather than Al6(Mn,Fe) phase.
3.3.2 Characteristic morphology of the secondary particleFigure 7 shows the embossed secondary particles in 3003 aluminum alloy obtained by deep-etching of Al matrix. In DC-cast sample, despite of deep-etching to a depth of around 50 µm from the initial surface, a full view of the particle was not detected. These script-like particles as shown in Fig. 7(a) were eutectic particles. In the near-surface region of HSTRC strip, secondary particles were observed along the cell/grain boundaries, as shown in Fig. 7(b) and (c). Homogeneous distribution of fine secondary particles were also observed inside of the cells. The particle exhibited spore-like morphology with a convex surface as show in Fig. 7(d). Motoi et al.11) observed similar α-AlFeSi phase nodules in the as-cast structure of commercial pure aluminum containing 0.13 Si and 0.13 Fe in mass% as impurities. They considered the formation behavior of the α-AlFeSi phase nodules as follows; the hydrogen gas dissolves in liquid of alloy during solidification. The hydrogen gas bubble acts as nucleation site of the α-AlFeSi phase. The α-AlFeSi phase nucleated at the inner-surface of the hydrogen bubble. It grows toward the center. Then, a spherical nodule of α-AlFeSi phase particle with a hollow shell is formed. In contrast to that, the spore-like particles in HSTRC strip in the present study exhibited various morphologies. Furthermore, the size of the particles is quite small; some particles are several hundred nanometers in diameter. These features do not match up with the formation behavior of nodules resulting from the hydrogen bubbles.
FE-SEM secondary electron images of deep-etched DC-cast sample and HSTRC strip. (a) Script-like eutectic particle, (b) HSTRC near-surface area, (c) particles at cell boundary, (d) spore-like inside of cell at near-surface.
In general, morphology of the crystal is controlled by its inherent crystal structure and growth conditions.12) The crystal structure of α-Al(Mn,Fe)Si is body-centered cubic with {110} planes as the highest reticular density. In fact, it leads to the growth of rhombic dodecahedron. This rhombic dodecahedron morphology of primary α-Al(Mn,Fe)Si phase has been confirmed by other workers.13,14) However, in our observation, the rhombic dodecahedron α-Al(Mn,Fe)Si phase was not found even in the strip central region. In 3003 aluminum alloy, the α-Al(Mn,Fe)Si phase is hard to be considered as the primary phase during solidification. Instead, the spore-like α-Al(Mn,Fe)Si phase formed from the trapped liquid droplet inside the cells. Thus, the spore-like particle was considered to be one of the eutectic components, not the primary phase. In this case, the cooling rate can strongly influence on the morphology rather than the crystal structure. Similar spore-like α-Al(Mn,Fe)Si phase particles were clearly observed in as-spun 3003 aluminum alloy ribbon fabricated by single-roll casting with much higher cooling rate than that of HSTRC.15)
3.3.3 Composition of the secondary particleIt was found that the solidification structure and secondary particles in the HSTRC 3003 aluminum alloy strip are quite different from those of DC-cast sample. In order to examine the detailed nature of the secondary particles in HSTRC strip, TEM analysis was conducted.
Figure 8(a) shows TEM bright-field micrograph of DC-cast sample. A diffraction pattern of the eutectic script-like particle corresponds to the [011] zone of orthorhombic Al6(Mn,Fe) phase (Fig. 8(b)). EDS analysis revealed that the particle contains 86.7 at% Al, 7.2 at% Fe and 5.4 at% Mn, close to Al6(Mn,Fe) as stoichiometric formula. Meanwhile, the α-Al(Mn,Fe)Si phase particles were not observed.
(a) TEM bright-field image of script-like particle in DC-cast sample, (b) selected area diffraction pattern of the particle in (a) corresponding on a [011] zone axis of Al6(Mn,Fe), (c) TEM-EDS spectra from the particle in (a).
Figure 9 shows a TEM bright-field micrograph of HSTRC strip. In the strip near-surface, particles are located along the cell boundaries as well as inside of the cells. The size of particles in HSTRC strip is much finer than that of DC-cast sample. Figure 9(b) shows a representative particle inside of the cell. It is the same as spore-like particles observed by FE-SEM. The small patches with different contrasts in the particle correspond to the convex surface pattern of the spore-like particle observed by FE-SEM (see Fig. 7(d)). From the selected area diffraction pattern of the particle inside of the cell, they are confirmed as body-centered cubic α-Al(Mn,Fe)Si phase. The particles on the grain boundaries are also α-Al(Mn,Fe)Si phase as shown in Fig. 9(c). It is of interest that the particles on the grain boundaries show the same orientation at specific observation condition. The secondary particles located at the grain boundaries are inter-connected along the grain boundaries (see Fig. 7(c)). From the results, it is assumed that these particles grew from a single nucleation event during solidification. In the strip central area, secondary particles are dispersed surrounding the α-Al grains as shown in Fig. 9(d). From the microstructure observation and XRD analysis found that these secondary particles are eutectic α-Al(Mn,Fe)Si phase.
(a) TEM bright-field image of HSTRC strip near-surface, (b) spore-like α-Al(Mn,Fe)Si particle inside of cell in HSTRC strip near-surface, (c) α-Al(Mn,Fe)Si particles on grain boundary in HSTRC strip near-surface, (d) eutectic constituents in HSTRC strip center area.
EDS analysis was also conducted to investigate the chemical composition of the secondary particles in DC-cast sample and HSTRC strip. The concentration of solute Mn, Fe, and Si elements with concentration of Al for the secondary particles was summarized as shown in Fig. 10. The EDS analysis results are consistent with those obtained by XRD and TEM analysis. In DC-cast, Si content is quite low and it is below 0.5% as shown in Fig. 10(a). These particles are considered to be Al6(Mn,Fe) phase. In contrast to that, secondary particles in HSTRC strip indicate high Si content. From the XRD and TEM observation results, these particles are α-Al(Mn,Fe)Si phase. For the α-Al(Mn,Fe)Si particles in the same HSTRC strip, the concentration of solute elements was different between strip surface and strip central area. Compared with the particles in the strip surface, the particles in strip central area show higher Fe concentration (Fig. 10(b) and 10(c)). In principle, Mn and Fe can substitute each other almost completely in the same α-Al(Mn,Fe)Si phase. However, the partition coefficient of Fe in Al is quite small compared with that of Mn. Also, the solubility of Fe in Al matrix is quite low. Therefore, the segregation of Fe into the liquid is high rather than that of Mn. In HSTRC, strip central area is considered as the final solidification part of the trapped solute-rich liquid. Thus, it can be assumed that relatively Fe-rich α-Al(Mn,Fe)Si phase was formed at HSTRC strip central area.
TEM-EDS results showing relative solute concentration ratio of the secondary particles in DC-cast sample and HSTRC strip. (a) DC-cast sample, (b) HSTRC strip surface, (c) HSTRC strip center.
In this study, solidification structure and secondary particles in the high-speed twin-roll cast 3003 aluminum alloy strip were investigated. The obtained results are as follows:
The authors acknowledge for the financial support from the Japan Foundry Engineering Society.
