Microstructural evolution and hardening phenomenon caused by aging of AlSi10Mg alloy by laser powder bed fusion

Abstract

Microstructures and age-hardening phenomena of directly aged (artificial aged) AlSi10Mg alloys fabricated by laser powder bed fusion (LPBF) were characterized using scanning transmission electron microscopy, atom probe tomography, and Vickers hardness testing. The microstructure derived from overlapping melt pools has a full cellular structure consisting of eutectic Si walls surrounding α-Al cells. In the initial stage of aging, solute clusters with density on the order of 10²⁴/m³ were formed in α-Al cells. By prolonging the aging time further, fine Si particles of about 50 nm in diameter precipitated. Before Si precipitation, the hardness of the aged sample was clearly greater than that of the as-built state. With further aging time, the hardness increased further because of the Si precipitation. Cluster analysis revealed that the number density and the size of clusters increased from as-built state by aging, whereas the types of the solute clusters remained almost unchanged by aging. The results indicate that the nanoscale clusters within the α-Al cells, which increase via aging, produce age-hardening effect.

1. Introduction

Laser powder bed fusion (LPBF), also known as selective laser melting, is a promising process to fabricate highly complex and flexible shapes from metal powders [1]. Reportedly, alloys fabricated by LPBF are given a higher cooling rate to solidify than by the conventional processes such as gravity die casting, thereby imparting higher strength [[2], [3], [4], [5], [6], [7], [8]]. High-strength and complex-shaped aluminum parts fabricated using LPBF are attractive to achieve weight reduction and higher fuel efficiency because aluminum alloys are used widely in many industrial fields such as aviation, automotive, military, and construction. Actually, because of its castability, weldability, low melting temperature, and low shrinkage, AlSi10Mg alloy is the most popular aluminum alloys for LPBF users [2]. Therefore, many investigations have been undertaken to clarify the unique microstructural characteristics, high strength, and the strengthening mechanism in the as-built state of AlSi10Mg alloy fabricated by LPBF [[9], [10], [11]]. The high cooling rate produced by the high-speed scanning laser irradiation with focused laser in the LPBF process engenders a highly supersaturated solid solution of Si in the α-Al matrix. The remaining Si forms sub-grain-like boundaries that crystallize as fine eutectic Si surrounding the α-Al cells. The amount of supersaturated solid solution of Si in α-Al can reach 2.2–2.8 at.% in AlSi10Mg alloy, which is much higher than the maximum solubility of Si of 1.4 at.% according to the equilibrium state of AlSi10Mg alloy [[9], [10], [11]]. Dimensional changes, corrosion resistance and high temperature strength derived from the characteristic microstructural morphology formed by the LBPF process have also been reported [[12], [13], [14]].

A solution heat treatment, a T6 treatment, and a direct aging (an artificial aging of as-built samples) used as heat treatment approaches for the AlSi10Mg alloy have been studied [[15], [16], [17], [18], [19], [20], [21], [22], [23]]. From many of those studies, decreased strength of AlSi10Mg alloy caused by heat treatment has been reported. The solution heat treatment on the AlSi10Mg alloy leads to decreased strength because of the coarsening of eutectic Si and because of decreased density of Si particles [15,19]. Subsequent aging at low temperature during T6 treatment also fails to contribute to strength; in fact, it reduces strength further [15,16,[19], [20], [21]]. Also, T5 treatment, by which AlSi10Mg alloys are aged at around 573 K without solution heat treatment, has been studied. The T5 treatment was originally intended for stress relief. This approach has been reported to soften the AlSi10Mg alloy while increasing its ductility [10,[18], [19], [20], [21]]. Some reports have described direct aging at a lower temperature than the standard T5 treatment. Several reports have explained that direct aging at around 473 K does not contribute to strength, although other reports have noted increased strength [[19], [20], [21], [22], [23]].

In our earlier work [9], we used scanning electron microscopy (SEM), electron backscatter diffraction (EBSD), scanning transmission electron microscopy (STEM) and three-dimensional atom probe (3DAP) to investigate characteristics of as-built AlSi10Mg alloy. The microstructure consists of the full-cellular structure in which the α-Al matrix was surrounded by eutectic Si. Results provided the quantitative values of supersaturated Si content obtained by high cooling rate and showed that solute clusters formed in the α-Al matrix because of the intrinsic thermal effect during fabrication. These microstructural features reveal that as-built AlSi10Mg via LPBF is about twice as hard as gravity casting with the same composition. These findings suggest two possibilities. The first possibility is that the basic microstructural morphology of overlapping melt pools produced by the laser irradiation greatly contributes to the high strength because of the eutectic Si walls surrounding α-Al cells and because of supersaturated Si in α-Al cells. The second possibility is that direct aging at a low temperature less than that of the T5 treatment after fabrication can achieve cluster strengthening and precipitation strengthening effectively because acceleration of the cluster formation using supersaturated solid solution and the precipitation using the existing solute clusters can be expected. In order to verify these possibilities, it is necessary to analyze the microstructural evolution due to aging of the samples. As described above, some approaches have been used for the quantitative evaluation of the microstructural evolution in the direct aging condition at low temperatures and its associated contribution to strength [[19], [20], [21], [22], [23]]. Nevertheless, these earlier reports have not described any microstructure characterization of the aged samples with respect to solute clustering.

For this study, we combine SEM, EBSD, STEM and 3DAP to precisely characterize the microstructures in the AlSi10Mg alloy. Details of the microstructural evolution and hardness changes of AlSi10Mg alloy during low temperature direct aging were investigated to explore ways to strengthen LPBF-fabricated AlSi10Mg alloy.

2. Materials and methods

The AlSi10Mg alloy powder used for this study was supplied by LPW Technology Ltd. The particle size values of D10, D50, and D90 respectively represent particles of 25, 41, and 66 μm in diameter. The specimens were prepared using powder having the chemical composition presented in Table 1. The processing conditions for the fabrication are shown in Table 2. The laser scanning strategy was designed using software (Magics; Materialise). The laser scanned alternately and the rotation angle of 90° between layers as shown in Fig. 1a. Rectangular specimens with length and width of 22 mm and height of 27 mm were fabricated (SLM280 HL; SLM Solutions) in high-purity Ar atmosphere (Fig. 1a). The temperature of the build base plate was maintained at 373 K during fabrication processes. After the fabricated specimens were divided into four sections, samples were taken from the same height of each sample, as presented in Fig. 1b. The samples were divided further and were evaluated either as-built state or after heat treatment state. All analyses were performed carefully at a position of 15 mm height from the bottom of the fabricated specimen. During the aging process, a dummy Al plate with an embedded thermocouple was placed in an electric furnace. After confirming stable conditions in which the range of 423 ± 2 K continues for several minutes or longer, the sample after being cut was placed on the dummy. The samples immediately rose in temperature because of heat transfer from the dummy. The dummy was kept precisely at 423 K while checking the temperature. The aging times chosen for direct aging were, respectively, 3, 5, 10, and 15 h. Samples aged for the prescribed time were immediately removed from the furnace and were then air cooled. Hereinafter, direct aging is abbreviated to DA, such a sample is designated as DA3h or as a similar term using the respective aging times.

Table 1.

Chemical composition of the A1Si10Mg powder used for this study (wt%).

Alloy	Si	Mg	Fe	Cu	Zn	Al
AlSi10Mg powder	9.9	0.34	0.11	<0.05	<0.01	Bal.

Table 2.

Parameters used for fabricating AlSi10Mg.

Laser power (W)	Scan speed (mm/s)	Hatch spacing (mm)	Laser spot size (mm)	Layer thickness (mm)
350	1150	170	80	50

Fig. 1.

(a) Fabricated sample and the scan strategy of AlSi10Mg and (b) schematic illustrations of the observation area.

The microstructures were characterized using optical microscopy (ECLIPSE LV150; Nikon Instruments Co.), SEM (JSM-7000F; JEOL), and EBSD (MSC-2200; TSL Solutions). The EBSD grain maps were taken from orientation imaging microscopy (OIM) data using image quality >1000 with 2 μm step size. Microstructural changes were observed using STEM (JEM-2100F; JEOL) equipped with an energy dispersive X-ray spectroscopy (EDS) detector. The samples for STEM were prepared using focused ion beam (FIB) machining (NB-5000; Hitachi Ltd.).

Atom probe tomography was executed on a local electrode atom probe (LEAP 4000X Si; Cameca SAS). The samples for 3DAP were prepared using FIB machining (Helios450; FEI Co.). The measurements were conducted in laser-pulsed mode with an evaporation rate of less than 0.1% ion/pulse. The condition for laser-pulsed mode were the following: laser pulse energy 25 pJ and the needle specimen temperature 50 K. Sufficient mass resolution for the analysis was achieved since experimental mass resolution of 370 (M/ΔM of the Al⁺ peak) full-width at half-maximum was obtained. Three-dimensional representation and a visual inspection of atom maps was performed to investigate the clustering search with the number density and the Mg/Si atomic ratio using IVAS software (Cameca SAS). The radial distribution functions (RDFs) analysis is applied to elucidate the interactions among constituent atoms of clustering [24]. The solute nearest-neighbor (NN) analysis with the maximum value of the d-pair distance (d_max) and the minimum size of the solute clusters (N_min) was also applied to clarify the distribution of solute atoms in the α-Al matrix.

Vickers hardness was measured using mirror finished samples at a 200 gf load and was determined using the average of five out of seven data points after excepting the upper and lower values. It is noteworthy that the as-built and all aged samples were measured for Vickers hardness at the same building height as that used for microstructural analyses.

3. Results

3.1. Microstructural analyses using OM, SEM, and EBSD

Optical micrographs, backscattered electron (BSE) images, and EBSD grain maps of the plane along the building direction (xz plane) and the parallel plane to the deposited layers (xy plane) in the samples of as-built and DA5h are shown respectively in Fig. 2, Fig. 3. The grain maps defined a grain boundary as a location with misorientation of more than 5°. The one laser scan track in the microstructure consists mainly of elongated columnar grains with 10–20 μm width and 40–100 μm length, as shown in the grain map of Fig. 2. In contrast, only the equiaxed grains were observed at the xy plane, as depicted in Fig. 3. They are divided roughly into two groups by grain size. One has large grains corresponding to the cross-section of the elongated columnar grains in Fig. 2. The others are fine ones formed at the boundary of the melt pool track. The enlarged views of the columnar regions are presented on the right side in Fig. 3a and b, as denoted respectively by the dashed squares on the left side BSE images of Fig. 3a and b. The single columnar grain is divided into fine α-Al cells with 400–600 nm in diameter by eutectic Si walls. Comparison with Fig. 2, Fig. 3 clarifies that the basic cellular microstructure in the samples of as-built remained unchanged even after direct aging at 423 K for 5 h.

Fig. 2.

OM micrographs, BSE images, and grain maps along the z direction, respectively, for (a) as-built and (b) DA5h.The BSE image in (a) is reprinted with permission from Heliyon 5 (2019) e01186, © 2019 Elsevier.

Fig. 3.

BSE images and grain maps perpendicular to the z direction, respectively, for (a) as-built and (b) DA5h. (c) Magnified BSE images of the right side are from the square area in left side images.

3.2. Microstructural analysis using STEM and 3DAP

A more in-depth microstructure characterization using STEM and 3DAP was conducted to clarify differences between the as-built and DA samples. Fig. 4 shows dark-field-STEM images and EDS elemental maps of Al, Si, Mg, Fe and O for the as-built and DA samples. In the microstructure of the as-built sample in Fig. 4a, fine primary α-Al cells are visible. The Si concentration at the outer perimeter of the α-Al cell is high because the outer area is the Al–Si eutectic region, although no elemental enrichment or segregation is observed inside the α-Al cell. In addition, many dislocations exist in the α-Al cells. The dislocations result from the large temperature gradient and rapid solidification which occurs during fabrication processes, as shown in Fig. 4a. These dislocations were reported from earlier studies of the α-Al cell of AlSi10Mg alloy, and other alloys using TEM/STEM analysis [[25], [26], [27], [28]]. Actually, DA3h cause no changes in the α-Al cells (Fig. 4b). No Si precipitation was observed inside most α-Al cells in DA5h. However, as shown by the arrows in Fig. 4c, there was an area of Si enrichment in one α-Al cell. The Si precipitates increased rapidly in the α-Al cell during the holding time of 10 h (Fig. 4d). Fe was enriched in some of the eutectic regions and is inferred to form the well-reported Al–Si–Fe compound [10]. O was also enriched in the eutectic region.

Fig. 4.

Dark-field-STEM images and EDS elemental maps of each element: (a) as-built sample, (b) DA3h, (c) DA5h, and (d) DA10h. Arrows on the maps of (c) and (d) indicate the Si precipitates.

Fig. 5 shows bright-field (BF)-TEM images and selected area electron diffraction (SAED) patterns obtained for DA3h and DA10h. In the BF-TEM image of Fig. 5a, dark-contrast dots are apparent in the α-Al cells. As shown in the elemental maps of Fig. 4b, no segregation of specific elements occurs in the α-Al cells of DA3h. In addition, the SAED pattern obtained from region A indicated by the yellow circle in Fig. 5a showed only FCC-Al (α-Al) reflection spot. Earlier works have shown that dot patterns reflecting clusters appeared in the α-Al cells of the as-built state [9,10,22]. As with these results, the dots in the α-Al cells of DA3h in Fig. 5a are not precipitates, but are probably solute clusters. In contrast, the SAED pattern taken from region B in the microstructure after aging time of 10 h (Fig. 5b) showed Si precipitation in the α-Al matrix.

Fig. 5.

TEM micrographs showing the aged samples: (a) DA3h and (b) DA10h. SAED patterns of α-Al matrix taken from the circle show the presence or absence of Si precipitates.

Fig. 6a shows a 3D atom map of DA5h. The Si-enriched region corresponds to a part of the eutectic region. To evaluate clusters and fine precipitates in the α-Al matrix and compared them with as-built state [9], a cluster analysis was applied in the cylindrical region (φ50 × 35 nm³) of the α-Al cell in Fig. 6a. Fig. 6b and c respectively present the RDFs of the normalized concentration for each atom around the Mg and Si atoms in the matrix. The analytically obtained results clarified that Mg atoms tend to exist around Mg atoms (Fig. 6b), whereas both Mg and Si atoms tend to exist to a slight degree around Si atoms (Fig. 6c). Fig. 7a shows the NN distribution of Mg atoms in the α-Al analyzed region depicted in Fig. 6a. In this figure “Random” means a natural distribution of the interatomic distances and “Observed” means the analytically obtained result of this work. In the present result, “Observed” is shifted to a lower NN distance than “Random”, which exhibits solute clusters in the α-Al cell. The d_max is defined as 1.25 nm as a result of Fig. 7a. The solute clusters portrayed in Fig. 7b are based on the cluster search algorithm using d_max = 1.25 nm and N_min = 5. The number of solute clusters in the cylindrical region is 77, which corresponds to a number density of approximately 1.1 × 10²⁴/m³. Fig. 7c shows the Mg/(Mg + Si) atomic ratio to the cluster volume. The ratio of Mg clusters to Mg–Si co-clusters was approximately 1:1.3, indicating that Mg clusters and Mg–Si co-clusters exist in almost equal numbers.

Fig. 6.

(a) 3DAP atom map for all elements in the sample aged at 423 K for 5 h (DA5h). An analyzed cylindrical region of φ 50 × 35 nm³ is included. Radial distribution functions around (b) Mg and (c) Si atoms, respectively, taken from the cylindrical region in (a).

Fig. 7.

(a) Solute NN analysis of Mg atoms in DA5h, (b) Atom map of Al, Mg and Si showing solute clusters, and (c) Mg/(Mg + Si) plotted as a function of the cluster volume. “Random” means a natural distribution of the interatomic distances; “Observed” means the analytically obtained result of this work.

3.3. Vickers hardness test

Fig. 8 shows the change in hardness of the as-built and DA samples. The hardness increased with aging time. It reached peak hardness at 10 h; then it decreased after 15 h. The behavior of the hardness change shows good agreement with the results reported for the time evolution of hardness reported earlier in the literature [21,22], although the reported aging temperatures differed: 448 K and 433 K, respectively. However, compared to those earlier studies [21,22], a different interpretation of the microstructural contribution to hardness change can be made. The interesting relation between hardness variation and microstructure evolution is discussed hereinafter.

Fig. 8.

Change in Vickers hardness of the AlSi10Mg as a function of aging time.

4. Discussion

Based on the results of microstructural observations presented in the preceding section, the microstructural evolution of as-built and subsequently aging samples can be summarized: 1) The microstructure derived from overlapping melt pools and the sizes of the grains in the as-built state remains unchanged after direct aging. 2) In the initial stage of aging (DA5h), solute clusters having density on the order of 10²⁴/m³ were formed in α-Al cells. 3) Aging at 423 K for 5 h is the critical condition for initiation of Si precipitation in the α-Al cells.

The hardness change in Fig. 8 is considered in association with the microstructural evolution described above. The hardness increases already in DA3h and DA5h before Si precipitation starts because of the high density of clusters (Fig. 7). Then, the hardness gradually reaches the peak holding time for 10 h because of fine Si precipitates (Fig. 4d). The Si size increases by prolonged aging and thus becomes incoherent with the FCC-Al matrix. Since incoherent Si particles make a smaller contribution to the strength, a decrease in hardness occurs in DA15h.

Solute clusters with a number density of 1.1 × 10²⁴/m³ were confirmed in DA5h, while our earlier work revealed a cluster density of 5.3 × 10²³/m³ in the as-built sample [9]. Consequently, the number density of clusters in the DA5h is an order of magnitude higher than that of the as-built state. It can therefore be inferred that the solute clusters contribute to the increased hardness. The mutual interaction of solute clusters and dislocations must be regarded as the strengthening by solute clusters. Solute clusters and dislocations interact according to two mechanisms: shearing (cutting) mechanism and Orowan bypass mechanism. Takata et al. reported that the yield strength increased because of the shearing mechanism in Al–Mg–Si alloy with Mg–Si co-clusters [29]. The authors presumed that solute clusters were cut by dislocations and were then dissolved into the Al matrix. Although the number density of the Mg–Si co-clusters was of the same order (10²⁴/m³) as that in the current study, the cluster size was not clarified in the literature. Fomin et al. proposed that a molecular dynamics model in which the transition of the two mechanisms (shearing and Orowan bypass mechanisms) of the solute cluster-dislocation interaction depends on the cluster size, strength, and deformation rate in Al–Cu alloy with Al–Cu co-clusters [30]. The authors concluded that the shearing and Orowan bypass mechanisms coexist depending on the size distribution of the solute clusters, but the size distribution threshold for different alloys remains unclear. Using the cluster analysis in Fig. 7 and our previous work [9], the changes in the cluster volume and the Mg/(Mg + Si) distribution of the number of clusters in the as-built and DA5h are shown in Fig. 9. Assuming a spherical cluster shape in Fig. 9a, the cluster radii of the as-built and the aged sample can be respectively as less than 1 nm and 1–2.3 nm. Accordingly, the increase in the number density and size of clusters by the direct aging contributing to age hardening at the initial stage of aging. Although the number and size of the constituent atoms of the solute clusters increased, the distribution of Mg/(Mg + Si) in all the solute clusters remained almost unchanged before and after direct aging, as shown in Fig. 9b. The underlying cause of strengthening by clusters requires further discussion, but at least the cluster analysis in this study demonstrates that the composition of the clusters did not change and that the increase in density and size simply resulted in strengthening by aging.

Fig. 9.

(a) Changes in the cluster volume and (b) the Mg/(Mg + Si) distribution of the clusters in the as-built sample and DA5h.

The first nearest neighbor (1NN) distance distribution was applied in the cluster analysis. However, it has been pointed out that the 1NN distribution of the clusters overlaps with the 1NN distribution of the matrix phase, leading to incorrect cluster selection when d_max is determined by the 1NN distribution [31]. Thus, we performed a cluster analysis of Mg atoms by considering KNN (K^th nearest neighbor) distance distribution. When K was increased, the d-pair distance of the clusters became slightly broadened, but the positional relationship between the analyzed d-pair peak and the natural distribution of interatomic distances did not change. Therefore, for the as-built and DA5h cluster analysis performed in this study, the 1NN distribution was found to be a more appropriate method than the KNN distribution.

As the main strengthening mechanisms for AlSi10Mg alloy fabricated by LPBF, Hall–Petch (because of eutectic Si surrounding α-Al cell), Orowan (because of Si precipitation), and dislocation hardening (because of high-density dislocation induced by the large temperature gradient and rapid solidification) have been proposed in the literature [[21], [22], [23],25]. Findings of the present study indicate that the nanoscale clusters within the α-Al cells produce age-hardening effects during the initial stage of aging. The greatest benefit of additive manufacturing including LPBF is the creation of a net-shaped three-dimensional structure. Consequently, strengthening achieved by solute clusters is an effective strengthening method for AlSi10Mg alloy via the LPBF process.

5. Conclusion

This study investigated the microstructure evolution and hardness changes of AlSi10Mg alloy fabricated by LPBF in as-built and subsequent direct aging at low temperatures to elucidate methods to strengthen LPBF-fabricated AlSi10Mg alloy. The basic cellular microstructure remains unchanged before and after direct aging, whereas nanoscale solute clusters and precipitated Si form within the α-Al cells by direct aging, thereby producing an age-hardening effect. In addition to the precipitation strengthening by fine Si, the strengthening effect of solute clusters was inferred because of enhanced interaction between dislocations and precipitates deriving from the increase in the sizes and numbers of the clusters. A new finding obtained from this study is that strengthening by solute clusters contributes to the hardening, as inferred from differences of cluster density and size, and from the period until Si precipitation in the α-Al matrix.

CRediT authorship contribution statement

Takashi Maeshima: Data curation, Writing – review & editing, Writing – original draft, Investigation, Conceptualization. Keiichiro Oh-ishi: Writing – review & editing, Formal analysis, Data curation, Conceptualization. Hiroaki Kadoura: Investigation.

Declaration of competing interest

The authors declare that they have no competing financial interests and personal relationships that could have appeared to influence the study in this paper.