Multimodal and multiscale investigation for the optimization of AlSi10Mg components made by powder bed fusion-laser beam

In recent years, there has been a growing interest in the use of additive manufacturing (AM) to fabricate metallic components with tailored microstructures and improved mechanical properties. One of the most promising techniques for the aerospace industry is powder bed fusion-laser beam (PBF-LB). This technique enables the creation of complex shapes and structures with high accuracy and repeatability, which is especially important for the aerospace industry where components require high precision and reliability. However, the impact of the PBF-LB process on microstructural features, such as the grain size distribution and porosity, remains an important area of research since it influences mechanical properties and performance of materials. In this study, a multimodal and multiscale correlative microscopy approach is used to investigate the microstructure of AlSi10Mg components made by PBF-LB. The study found that the correlative microscopy approach involving X-ray images with visual, chemical, and diffraction information coming from optical microscopy (OM), scanning electron microscopy (SEM) and electron backscatter diffraction (EBSD) is highly effective in reaching a more comprehensive understanding of the relationship between the fabrication process and the effective microstructure of PBF-LB fabricated components enabling the optimization of their performance for a wide range of applications.


Introduction
Additive manufacturing (AM) has revolutionized the way metallic components are fabricated.In recent years, there has been a growing interest in using AM techniques to fabricate metallic components with tailored microstructures and improved mechanical properties [1,2].One of the most promising AM techniques is powder bed fusion-laser beam (PBF-LB), broadly referred to as Selective Laser Melting (SLM), which enables the creation of complex shapes and structures with high accuracy and repeatability [3].PBF-LB involves the use of a high-power laser to fuse metal powder into solid components.The process begins with the creation of a 3D model of the component using computer-aided design (CAD) 1 3

Sample preparation
Two metallographic samples were prepared for each test piece, with one section being perpendicular and the other parallel to the cylinder's direction of construction (z-axis).All sections obtained from the coupons and test pieces were embedded in Multifast black phenolic resin (Struers), and polished using 1 mm diamond paste (DP-Suspension, Struers).The microstructure of the aluminum alloy was revealed using an etching solution containing Keller's reagent (2.5%v/v HNO 3 , 1.5%v/v HCl, 1%v/v HF, 95%v/v H 2 O).

Optical microscopy (OM)
OM examination was carried out using a Leica M205C and Leica DM6000M optical microscopes and a Hirox KH-7700 digital microscope.

Image analysis and processing
Once the XRM datasets were reconstructed, Dragonfly Pro (Version 2022.1)from Object Research Systems [25] was employed to filter and process the images.

Optical and scanning electron microscopy analysis
A CAD representation of the AlSi10Mg sample fabricated via PBF-LB with a density of 99.4 ± 0.1% is reported in Fig. 1a.The cut plane orthogonal to the construction axis (red) and the parallel one (blue) are highlighted.A metallography obtained from the orthogonal section (red) with respect to the construction axis is shown in Fig. 1b.The traces (hatch) left by the laser scan can be distinguished as they are oriented along bundles of parallel lines, but with different angles between layers.The traces are parallel in a single layer and rotated, typically at an angle of 67°, when passing to the next layer to maximize the number of orientations, thus mitigating the anisotropy problems of these materials [26].The width of the trace corresponds to the width of the melt pool, as will be extensively discussed later.The average width of the traces is 170 ± 30 microns, based on a minimum of 10 measurements.The entire surface of the specimen appears to be free of obvious defects, with sporadic almost spherical black areas, associated with holes generated when a vapor-filled cavity is formed by the laser beam, generally referred to as keyhole in literature [27,28].
Another type of defect found on all specimens consists of irregularly shaped particles, as shown in Fig. 1c compatible in shape and size with the presence of oxides, as reported in the literature [29].
An enlarged view of a keyhole obtained using a digital microscope revealing a height of 15.22 µm and a width of 28.72 µm is depicted in Fig. 1d.Observing Fig. 1e, it is evident how the edge of the specimen is significantly different from the core, being composed of overlapping coincident traces, no longer phased by 67°.The presence of non-continuous traces in the upper layers, as observed in Fig. 1b, depends on the imperfect perpendicularity between the construction axis of the specimen and the sectioning plane, as well as on variations in the depth of the trace during the fabrication process.The situation is depicted in Fig. 2 where a depth variation of the laser scan is reported for the purpose of visualization.Note that the depth variation in Fig. 2 is stressed and out of the scale, hence should be considered just as an explanation of the phenomenon.A close-up of a region where scanning traces intersect is shown in Fig. 3, with the Fig. 1 a CAD representation of the sample AlSi10Mg fabricated via PBF-LB with a density of 99.4 ± 0.1%.The cut plane orthogonal to the construction axis (red) and the parallel one (blue) are reported.b A metallography obtained from the orthogonal section is depicted.The traces (hatch) left by the laser scan, with an average width of 170 ± 30 µm, can be distinguished revealing a rotation of 67° between layers and pores (keyholes) are highlighted by yellow circles.c The specimen is characterized by the presence of irregularly shaped particles which can be associated with the existence of oxides according to their geometry and size.d Enlarged view of a keyhole obtained using a digital microscope revealing a height of 15.22 µm and a width of 28.72 µm.e The edge of the specimen is significantly different from the core, being composed of overlapping coincident traces, no longer phased by 67° intersections highlighted in red dotted lines.It is evident that the edge of the single scan is not continuous, but rather composed of a cellular structure, as clarified in the following SEM investigation.
The average hardness value of the samples, obtained using the Micro Vickers method (ASTM E384), is 140 ± 5 HV.The applied load was 30 N, and the reported result is the average of at least 50 measurements.This value is significantly higher than that of the same alloy obtained by casting in the annealed state, which ranges from 95 to 105 HV [29].The typical microstructure of the AlSi10Mg alloy obtained by PBF-LB is shown in Fig. 4a: a solid solution of aluminum-alpha (Al-α), depicted in black, surrounded by a continuous network of Al-Si eutectic, represented in white.The elemental composition of the eutectic, obtained through EDS, is reported in Fig. 4b.The silicon content is slightly above the nominal 10% revealing that the faster cooling results in a supersaturated pseudo-eutectic microstructure [28].The microstructure observed Fig. 4a is consistent with the cellular structure, as defined in the literature [30].A schematization of the PBF-LB fabrication process is represented in Fig. 5a.At the center of the scan trace, the material undergoes a cooling process to the ambient temperature, resulting in the formation of a fine cellular structure.The production process of the component involves slightly overlapping each scanning trace with the previously deposited one to maximize fusion.This technique results in a portion of the already deposited material experiencing a second temperature increase like that occurring during welding.This process initially leads to the coarsening of the structure, followed by the formation of a heat affected zone (HAZ).The HAZ is characterized by the disruption of the continuous eutectic network, resulting in the formation of a disordered structure, as shown in Fig. 5b [31].The region observed using SEM is reported at lower magnification in the OM image, in Fig. 5b.
The visible traces observed via optical microscopy in the section parallel to the construction axis, highlighted in blue in Fig. 1a and reported in Fig. 6, also show a change in microstructure, with a transition from fine to coarse cellular structure and the presence of a thermally altered overlapping zone.In the micrograph of Fig. 6, the directionality of heat transfer is noticeable: the cellular microstructure in the fusion pools is not equiaxed but presents an elongation towards the border of the pool, which is the direction of maximum heat removal.
A Building Direction (Z0) Inverse Pole Figure (IPF) map obtained via EBSD is presented in Fig. 7.The image proves the smaller size of the grains in the pool border compared to the center.This observation is consistent with the previous statements.Despite the increase in size of the matrix in the cellular structure towards the melting pool border, the grain size decreases, as illustrated in Fig. 7.It is noteworthy that nearly all grains exhibit an equiaxed shape, as anticipated for a section perpendicular to the building direction, where isotropic heat diffusion occurs in this plane.
Fig. 2 The presence of noncontinuous traces depends on the imperfect perpendicularity between the construction axis of the specimen and the sectioning plane, as well as on variations in the depth of the trace during the manufacturing process.The depth variation reported in this figure is stressed and out of the scale, hence should be considered just for the purpose of visualization However, the exception is observed in the smaller grain present in the border which exhibits an elongated shape.This elongated shape confirms the occurrence of strong heat diffusion through the border of the melting pool [32].Upon observing the EDS map of the same area, as presented in Fig. 8, it is evident that the defects in the Al-Si matrix (depicted by the black areas in the Al and Si maps) are primarily composed of O, C, and Fe.This confirms the presence of oxide defects, as discussed earlier.Furthermore, the Si map illustrates the scan tracks distinctly.Si has a noticeable difference in melting temperature compared to Al, and this gradient is the cause of the slight variation in Si distribution among the samples.This is due to the segregation mechanism, a well-known effect in literature [33].
Two subsets of the IPF map from Fig. 7 are reported in Fig. 9 where grain sizes smaller and larger than 7 µm are considered separately.The reason for this subdivision is that a preferential orientation of the grains is observed, and 7 µm is found to be the optimal value that highlights this feature.It should be noticed that 2 µm is the smallest size that can be resolved with the current resolution, where the step size is 0.5 µm.The smaller grains (2 µm < Ø < 7 µm) demonstrate a 101 preferred orientation in the build direction, whereas the coarser grains (Ø > 7 µm) show a 001 preferred orientation.No preferred orientations are observed in the x and y directions.6 The micrograph related to the section parallel to construction axis revealed a change in the microstructure, with a transition from fine to coarse cellular structure and a thermally altered zone.This micrograph highlights the directionality of heat transfer: the cellular structure in the fusion pools is not equiaxed but shows an orientation towards the border of the pool, which is the direction of maximum heat removal Fig. 7 Inverse pole figure (IPF) map in the building direction (Z0), normal to the plane of the image.It is worth noting that the grains at the border of the pool exhibit a smaller size compared to those in the center Fig. 8 Energy dispersive X-ray spectroscopy (EDS) map of the same area shown in Fig. 7 Fig. 9 The preferred crystallographic orientation exhibited by small grains (2 μm < Ø < 7 μm) in the building direction is 101, while coarser grains (Ø > 7 μm) show a preferred orientation of 001.No preferred orientation is observed in the x and y directions of interest, called VOI 2. The pore spaces from VOI 1 and VOI 2 were transformed into a multi-ROI to perform an object analysis and extract the pores mean Feret diameter.The Feret diameter is measured by selecting the two points on opposite sides of the object's boundary that are farthest apart and measuring the distance between them in various directions.The mean Feret diameter is then calculated as the average of all these measurements.Therefore, this parameter is particularly suitable in applications such as pore size analysis.
A comparison between results obtained for VOI 1 and VOI 2 is reported in Fig. 12 as well as the blue cylinder that highlights the internal volume of the sample that we called VOI 2. Observing the mean Feret diameter distribution histograms reported in Fig. 12a-b referred to VOI 1 and VOI 2 a significant variation in the mean value from 74.52 µm to 23.65 µm emerges.These results suggest further XRM experiments to reach a higher resolution and investigate the pores behavior with finer detail.Hence, a third volume of interest, VOI 3, was selected to be scanned using a pixel size of 1.51 µm.The same analytical approach was used for segmentation and object analysis.In this case, the mean value of the mean Feret diameter distribution referred to VOI 3 is 10.60 µm, as shown in Fig. 13.This behavior can be explained as follows.When using XRM to analyze the porosity of a material, the size of the pixels used in the scan can have a significant effect on the measured pore size.When using a larger pixel size of 16.82 µm, the image of the sample would contain fewer pixels and each pixel would represent a larger area of the sample.As a result, small pores would be smoothed out and merged with neighboring pores, leading to an overestimation of the pore size.On the other hand, large pores would be more accurately represented, and their size would be less influenced by pixel size [37].This could explain the larger mean value of the mean Feret diameter of pores observed with a larger pixel size.When using a smaller pixel size of 3.62 µm or 1.51 µm, the XRM image would contain more pixels and each pixel would represent a smaller area of the sample.This would enable the detection of smaller pores, which would lead to a more accurate measurement of pore size, especially for small pores.As a result, the measured mean value of the mean Feret diameter of pores would decrease as the pixel size decreased.
In summary, the variation of the mean value of the mean Feret diameter of pores with changing pixel size in X-ray microscopy scans is due larger pixel sizes tend to overestimate pore size while smaller pixel sizes tend to provide a more accurate measurement of pore size, especially for small pores.
XRM results highlight the presence of a matrix of pores that evolves at different length scales which can significantly affect the mechanical properties of the sample.Porosity can reduce the strength, stiffness, and ductility of the material, as well as increase its susceptibility to fatigue and fracture [38][39][40][41][42][43].The larger the pore size, the more significant the reduction in strength and stiffness [41,44].However, small pores can also play a role in reducing strength and stiffness by acting as stress concentration points [45].The pore matrix dataset can be used to inform different damage models as in [46,47] and implemented in a set of global finite element simulations.By simulating the mechanical behavior of a representative volume element of the material, the effect of the evolving matrix of pores on the mechanical properties of the material can be investigated.The simulation results can be used to optimize the properties of materials for specific applications by understanding how porosity affects the mechanical behavior of the material.

Conclusions
In this study, a multimodal and multiscale correlative microscopy approach was employed to investigate the microstructure of AlSi10Mg components fabricated by Powder Bed Fusion-Laser Beam (PBF-LB), a promising additive manufacturing technique for the aerospace industry.The study aimed to provide a comprehensive understanding of the relationship between the fabrication process and the effective microstructure of the components involving a detailed exploration of each investigation method, highlighting their unique characteristics, providing evidence of their positive aspects, and acknowledging the limitations of other methods.Optical Microscopy (OM) was used to investigate the traces (hatch) left by the laser scan revealing that they are oriented along bundles of parallel lines, with a rotation angle of 67° between layers to mitigate anisotropy issues.The average width of the traces was measured to be 170 ± 30 microns.The presence of irregularly shaped particles, likely oxides, was also observed.Scanning Electron Microscopy (SEM), coupled with Energy Dispersive X-ray Spectroscopy (EDS), was used to characterize the chemical composition and microstructure of AlSi10Mg alloy highlighting the transition of the matrix of the cellular structure from fine to coarse zones and the presence of a Heat Affected Zone (HAZ).Electron Backscatter Diffraction (EBSD) showed that the grains at the border of the pool exhibit a smaller size compared to those in the center.EBSD revealed a preferred crystallographic orientation for small (101) and large (001) grains.Then, multiscale X-ray

Fig. 3 Fig. 4 aFig. 5 a
Fig. 3 Red dotted lines highlight a region where scanning traces intersect.It is evident that going from the center to edge of the scan trace the matrix of the cellular structure increases in size

Fig.
Fig.6 The micrograph related to the section parallel to construction axis revealed a change in the microstructure, with a transition from fine to coarse cellular structure and a thermally altered zone.This micrograph highlights the directionality of heat transfer: the cellular structure in the fusion pools is not equiaxed but shows an orientation towards the border of the pool, which is the direction of maximum heat removal

Fig. 11
Fig.11The selected VOI is delimited by the red cylinder (left).The region of interest (ROI) containing the pore space was segmented using a histogram-based thresholding (right)

Fig. 12 Fig. 13 a
Fig. 12 Comparison between mean Feret diameter distributions related to the pores from (a) VOI 1, delimited by the red cylinder (pixel size 16.82 µm), and (b) VOI 2 highlighted by the blue cylinder (pixel size 3.62 µm).A significant variation in the mean value from 74.52 µm (VOI 1) to 23.65 µm (VOI 2) emerges.This suggests performing further experiments to reach higher resolutions

Table 1
Key characteristics of AlSi10Mg metal powder for PBF-LB fabrication: tap density, flow rate, and particle size distribution, aligned with company purchase specifications, as provided by the powder supplier

Table 2
Process parameters and build chamber atmosphere specifications SEM investigation was conducted using a field emission gun scanning electron microscope (FE-SEM) (Ultra-Plus, Carl Zeiss, Oberkochen, Germany) equipped with an Energy Dispersive X-ray Spectroscopy (EDS) detector (INCA X-EDS Oxford Instruments United Kingdom) and an Electron Backscattered Diffraction (EBSD) detector (C Nano Oxford Instruments United Kingdom), using a 0.5 μm scanning step size.FE-SEM analysis samples cleanup was carried out in ultrasonic bath with Methyl Acetate for 20 min at 40° C. Methyl Acetate (98%) was purchased from Sigma Aldrich.XRM scans were performed using a ZEISS Xradia Versa 610 available at the Research Center on Nanotechnology Applied to Engineering of Sapienza (CNIS) that is part of the open Infrastructure for "Advanced Tomography and Microscopies (ATOM)" of Sapienza University of Rome.The first low resolution scan was conducted to obtain an overall understanding of the porosity throughout the structure.Subsequently, the Scout-and-Zoom approach was employed to obtain three additional scans, each with a smaller pixel size, to study the porosity at different length scales.