Analysis of the Topographical, Microstructural and Mechanical Surface Properties of Powder Bed Fusion Melted AlSi10Mg for a Broad Range of Process Parameters

Abstract

The topographical, microstructural, and mechanical surface properties of additively manufactured components depend on variations of several processing parameters. Most studies focus on a narrow range of parameter variations, with the surface and subsurface characteristics being determined for that limited set of conditions. This makes it difficult to optimize these properties for additively manufactured parts and the energy consumption of the additive-manufacturing (AM) process. Our study looks at the systematic variation of two key AM parameters over their full range using a commercial AM machine. The laser scanning speed (500–1700 mm/s) and the laser power (250–370 W) were the parameters used. We analyze and discuss how these two parameters affect the surface topography, roughness, porosity, microstructure and hardness, as well as their anisotropy for the top and side surfaces during powder bed fusion, using a single AM machine and printing strategy. The aluminum alloy AlSi10Mg was selected for the study. It is one of the most commonly used materials in die casting and has the potential to take advantage of AM technology, since these parts can be lightweight, have good mechanical properties and to be produced with complex shapes.

1. Introduction

With the need to save energy and minimize carbon emissions, additive manufacturing (AM) [1], especially selective laser melting (SLM) [2,3], is becoming increasingly important in manufacturing. Advances in AM over the last decade have seen it evolve from a prototyping technology to the point where it is transforming the way products are designed, developed and manufactured, and with improved properties and performance [4,5]. However, to achieve this, several parameters of the AM process need to be set correctly. The most important ones are the scanning speed [6] and the power [2,3] of the laser beam, which lead to a specific melt-pool geometry as well as generating and influencing the temperature gradients, solidification rates, dendritic microstructures and the wetting behavior [4,7,8]. Of course, these properties also depend on the scanning strategy [9], layer thickness [9,10] and hatch spacing [11].

The variation of these AM processing parameters and the use of different materials leads to printed metal components with unique (1) surface properties in terms of topography, roughness, texture, morphology, as well as (2) microstructure, porosity, and (3) mechanical properties such as hardness. All three groups of properties, which are mainly related to the surface and subsurface, are very different from the properties generated by conventional methods [7,12], as explained below.

AM produces surface features (1) with a very specific topography and texture on the outer surface, which typically have a greater roughness than when using conventional methods. However, the surface roughness levels on the top and side surfaces can be reduced by tailoring the AM process parameters, e.g., by optimizing the laser power, scanning speed and hatching distance [13]. It was reported that at larger energy densities the formation of defects, cracks and pores occurs, while the balling effect predominates at smaller densities [1]. In addition, the hatch distance can also strongly influence the topography, where narrow scan tracks cover the surfaces evenly. However, due to the large number of different AM processing parameters, laser characteristics and AM machines, it is difficult to capture and explain all the impacts on the main topographical features.

In addition, most studies use only the most common parameter Ra to describe the surface roughness or topography [13,14,15,16]. However, other roughness parameters, such as Sq, Sdq, Ssk, Sku and volumetric parameters, are often more relevant for specific application requirements and define the main functionality of components [9,11,17]. Various surface-roughness parameters (Sa, Sq, Sku) were investigated in [18,19], but no effects of the printing parameters were investigated.

Another challenge in the correlation between the AM parameters and the resulting functional surface properties is the typical mechanical post-processing, in which the surfaces are prepared with a much finer roughness than result from the AM. This means that the original AM surface topography is removed, but the area below the surface still contains the original (2) porosity and microstructure—another important (sub)surface property that greatly depends on the AM process parameters. For example, it was shown in [20] that the mechanism of pore formation in the melt pools formed by the laser-melted track depends on where the melt pores are formed, at the boundaries, and where the gas pores are formed, inside the melt pools. However, only one set of process parameters was used in that investigation.

While pores are always present to a certain degree in AM, the density of the printed parts can reach a value of 99.9% under optimal processing parameters [21]. But as mentioned above, these optimal processing parameters for reducing the number of pores can be different for different laser types, AM machines and materials. When trying to adjust the processing parameters for the desired surface roughness and porosity, both properties and their relationships to the processing parameters must be known.

Finally, the mechanical properties of the surface (3) are also a crucial surface property that is directly influenced by the AM processing parameters. They are very sensitive to the microstructure, including phase constituents, grain size and morphology, dendrites and elemental distribution. In tribological applications, surface hardness is often one of the most important mechanical properties for wear resistance [22,23]. In this type of AM process, the low melt volume results in extremely high cooling rates of the molten metal [24], which leads to a typical fine dendritic, eutectic, cellular microstructure of Al and Si, resulting in a higher hardness than conventionally produced materials [25,26,27,28]. In addition, different regions of the microstructure occur within the melt pool due to the different thermal history [26,29]. Accordingly, extensive knowledge about the relationship between the microstructure and the hardness has already been generated. However, a direct relationship between the processing parameters and the size of the microstructure and its effect on hardness was not yet reported, but this is crucial for predictable AM for a desired specific hardness.

To summarize, each AM component will have specific topographical, microstructural and mechanical surface properties depending on the variation of the processing parameters. In addition, it is also a function of the build direction due to gravity. Therefore, the properties will change in different orientations of the part. However, most previous studies focused on a limited number of specific AM parameters. Accordingly, a comprehensive understanding of the effects that the AM process parameters have on the functional properties of the surfaces and their subsurface area is still lacking. This affects the possibility of optimizing these properties for the functional optimization of AM parts and the energy consumption of the process.

The aim of our investigation was to use a systematic and comprehensive set of AM process parameters and relate them to the resulting topography, roughness, microstructure and hardness in relation to the printing direction of an aluminum alloy, but performed under the same conditions, with a single AM machine and printing strategy and for a wide range of laser powers and scanning speeds. The aluminum alloy used was AlSi10Mg, which is one of the most widely used materials for die casting, with the potential to take advantage of AM technology, since the parts need to be lightweight with complex shapes and have good mechanical properties [30].

2. Materials and Methods

2.1. Material and AM Processing Parameters

We used certified aluminum AlSi10Mg alloy powder (EOS, Munich, Germany) [31] with the composition 9–11% Si, −0.55% Fe, −0.05% Cu, −0.45% Mn, 0.25–0.45% Mg, −0.05% Ni, −0.10% Zn, −0.05% Pb. −0.05% Sn, −0.15% Ti (the rest is Al), which was carefully handled and stored to prevent moisture absorption or contamination. The particle size distribution of the spherical powder was in the range 25–70 μm with an average of 50 μm [31]. The laser powder-bed fusion (LPBF) equipment was an EOS M290 (Munich, Germany), which uses a Yb fiber laser with a wavelength of 1060–1100 nm. The laser operates with a Gaussian (TEM₀₀) intensity distribution, meaning energy is highest at the beam center and gradually decreases toward the edges. The printing chamber was pre-filled with argon to minimize the presence of oxygen. Two sets of sixteen cubic specimens (all surfaces must be flat to perform the measurements, Figure 1b) with dimensions of 10 × 10 × 10 mm³, as shown in Figure 1a, were printed with the widest reasonable range (based on preliminary experiments) of the two main processing parameters, i.e., laser power P (250–370 W) and scanning speed v (500–1700 mm/s). The hatch distance h was set to 190 μm, and the layer thickness d of 30 μm was constant for all the specimens. The powder bed was layered under industry standard conditions to ensure high-quality layer deposition. The recoater system—comprising a smooth ceramic blade and nickel roller, guaranteeing smooth and uniform powder spreading for each layer. The calculated energy density η from Equation (1) corresponds to the wide range of 26–130 J/mm³. The varied most important parameters of the AM process are listed in

Table 1. Processing parameters of printed specimens and their energy densities.

Sample No.	Laser Power [W]	Scanning Speed [mm/s]	Energy Density, η [J/mm3]
1	250	500	88
2	250	900	49
3	250	1300	34
4	250	1700	26
5	290	500	102
6	290	900	57
7	290	1300	39
8	290	1700	30
9	330	500	116
10	330	900	64
11	330	1300	45
12	330	1700	34
13	370	500	130
14	370	900	72
15	370	1300	50
16	370	1700	38

.

$$\eta ={\displaystyle \frac{P}{v h d}}$$

(1)

The process starts with printing a grid mesh, with a height of 3 mm, on top of the base plate. After the mesh is printed the specimen is printed on top of the mesh with the chosen processing parameters. The purpose of the grid is to facilitate the removal of the printed specimen from the base plate and not to damage the top and side surfaces. The other processing parameters and settings were: chessboard scan strategy with 67° rotation of the scanning direction for each printed layer (Figure 1c), alternating scan directions of the scan tracks, 80 μm laser spot size and 200 °C preheating of the base plate.

To simplify the interpretation of the large parameter space (laser power and scanning speed) and its corresponding energy density, the experimental conditions are grouped into four distinct quadrants based on their relative energy density levels. This approach allows for a more systematic comparison of trends across the dataset and highlights the relative influence of energy density on the observed outcomes, rather than analyzing individual parameter combinations. With this grouping, we also want to facilitate clearer discussions of trends, such as increases or decreases in energy density, while still capturing the broader behavior within each quadrant. The 4 quadrants are as follows: (1) Quadrant I–specimens are produced with a relative moderate energy density (calculation and values are reported in Table 1), with a low laser power (250, 270 W) and a low scanning speed (500, 900 mm/s), (2) Quadrant II–specimens are produced with the lowest energy density, with low laser powers (250, 270 W) and high scanning speeds (1300, 1700 mm/s), (3) Quadrant III-specimens are produced with the highest energy density, at high laser powers (330–370 W) and low scanning speeds (500, 900 mm/s), and (4) Quadrant IV-specimens are produced with a moderate energy density, but here instead of a low power, as in quadrant I, a high laser power is used (330–370 W) with high laser speeds (1300–1700 mm/s).

2.2. Topography, Microstructure and Hardness Analysis

The surface topographies of the specimens from the top and side surfaces were measured with an optical 3D microscope from Bruker-ContourGT-K0 (Tuscon, AZ, USA), 2.5× magnification, a surface topography area of 1.7 × 1.3 mm² with a lateral resolution of 1.34 μm. For each energy density, on both duplicate specimens, six measurements were taken on each top and side surface to obtain an average value and all the roughness parameters were calculated for each specimen. The following areal texture parameters were considered according to ISO 25178-2 [32]: Sa (arithmetic mean of the height of a line), Ssk (skewness), Sku (kurtosis), which describe the probability density of the sharpness or asymmetry of the surface, and Sdq (root-mean-square gradient), which defines how smooth the surface is. The surface roughness parameter Sdq is calculated as the square root of the gradients at all the points of the defined area, i.e., if a surface is completely flat, the value of Sdq is 0. A higher value of this parameter corresponds to a surface with larger gradients and a surface that is less smooth. A surface with a slope component of 45 degrees has a value of 1. These 3D roughness parameters were calculated and analyzed using the MountainsMap imaging topography V8 software (DigitalSurf, Besançon, France). The surface data was post-processed to calculate the roughness parameters. The F operator to remove the surface slope consisted of subtracting a mean plane using the least-squares plane method to remove the overall surface slope and filtering with a noise filter λ_S of 25 μm and a roughness filter λ_L with a cut-off of 0.8 mm.

To gain a better insight into the surface integrity, microstructure analyses were carried out. Both sets of specimens were first ground using grinding papers of p400, p800, p1200 and p4000 grades, and then polished with a 3-μm diamond suspension on a cloth. The polished surfaces were then etched with Keller’s reagent using 2 mL of HF (hydrofluoric acid), 3 mL of HCl (hydrochloric acid), 5 mL of HNO₃ (nitric acid) and 190 mL of distilled water for 16 s. One set of specimen’s surfaces were analyzed in detail using a 3D digital optical microscope (Hirox HRX-01, Besançon, France) with 30× magnification at a 5-μm lateral resolution for the porosity analysis and 140× magnification at 1-μm lateral resolution for the microstructure analysis. The porosity images of the polished surfaces were then processed with a special Hirox software (Hirox HRX-01, Besançon, France) module for the particle analysis to determine the porosity values of the surface.

To obtain additional information about the formation of the surface features and changes in the microstructural areas, selected specimens were examined using scanning electron microscopy (JEOL JSM IT100, Tokyo, Japan) in conjunction with an EDS (energy-dispersive spectroscopy) detector (JEOL JSM IT100, Tokyo, Japan) to analyze the chemical composition, the distribution of alloying elements and the presence of an oxide layer.

To test the mechanical properties, the microhardness on the polished surface was measured with a Vickers indenter (Leitz-Miniload 2, WILD Leitz GmbH, Wetzlar, Germany) by applying a load of 500 g for 30 s. On each specimen of duplicate processing parameters, seven measurements were performed on each specimen and the average value was calculated, with a standard deviation.

3. Results

3.1. Topography

Figure 2 shows the topography of the top surfaces for the individual processing parameter of the respective laser power and laser scanning speed. In general, the surfaces become increasingly flat with a higher energy density (e.g., with increasing laser power and decreasing scanning speed). The height difference in the topography changes from 140 µm to 40 µm. Increasing the laser-scanning speed results in the topography having more textured, elongated lines, while increasing the laser power reduces the height of these elongated lines.

The greatest difference in the topography, which is due to the greatest difference in the energy input, was observed for quadrants II and III. In quadrant II, the height difference of the surface reaches an overall maximum of 140 µm. The specimens show the most pronounced texture, which consists of elongated, parallel height peaks. In quadrant III, the height difference of the topography was the smallest. The surface topography appears relatively flat, with a height difference of up to 60 µm. The surface has a certain texture in the form of ripples. These ripples are curved, have different orientations and are more clearly visible at 300 W and 500 mm/s. At 370 W and 500 mm/s, a minimum height difference of 35 µm was achieved. At a moderate energy density in quadrant I, the surfaces have an average height difference of about 80 µm and the elongated surface texture is reduced by smoothing these peaks. Accordingly, in quadrant IV, the surface texture of the elongated contour lines is again more predominant, but the height difference is still medium at about 80 µm.

Figure 3 shows the topography of the side surfaces for each laser power and laser scanning speed. The side surfaces are rough, fragmented and there are no large areas with the same surface height. The difference in height gradients is large, with the range between the lowest and highest points being approximately 170 µm. As the scanning speed increases, the surfaces show more circular elevations, with the protrusion on the surfaces becoming higher and sharper as the laser power increases. In terms of build direction, there are no significant differences in the orientation of the surface, except at 250 W and 1300 mm/s where there is a line of increased surface height that is perpendicular to the build direction.

In quadrants II and IV, the height distribution of the surface becomes more even, with less height gradient. Especially at higher laser scanning speeds, there is a large number of small, circular height areas. In quadrants I and III, there are large, fragmented areas with a height increase of up to 180 µm (achieved at 370 W and 500 mm/s). Between these areas, there is usually a relatively flat surface.

Figure 4 shows cross-section profiles for the printed surfaces. Selected processing parameters (power, speed) represent the boundary parameters from the matrix to show the differences and trends more clearly. None of the surfaces are flat; they exhibit a distinct periodic waveform that varies in amplitude and wavelength.

When the samples are made with the lowest laser power of 250 W and a scan speed of 500 mm/s (Figure 4a), the extracted profile consists of a larger wavelength profile with some additional smaller waves in between. The smaller waves have a wavelength of about 190 µm, while the larger ones have a wavelength of 600 µm, with a height amplitude of 50 µm. On these surfaces, some of the upper surface peaks are removed and flattened, as you can see in the optical image. At higher scanning speeds (250 W, 1700 mm/s), where the energy density is lowest (Figure 4b), there are several thin peaks (the upper radius is large) with an increased height of 30 µm and a wavelength of about 190 µm. At the highest energy density (370 W, 500 mm/s), shown in Figure 4c, the peaks are extremely flat, with a wavelength of 190 µm and a large peak radius. Due to the amplitude of less than 10 µm, some of the peaks begin to merge. With increasing laser power at high scanning speeds (370 W and 1700 mm/s), as shown in Figure 4d, the peaks are more homogeneous with a large top radius, a smaller amplitude of 20 µm and uniform spacing of 190 µm.

Figure 5 is an SEM image of an area of the upper surfaces and their features for the selected processing parameters (power, speed), which represent boundary parameters from the matrix. All the images show two adjacent laser tracks that are 190 µm apart, with the boundary between them indicated by the red dashed line. For each processing parameter, there are changes in the features that define their surface properties.

In Figure 5a, at a low scanning speed and power (250 W and 500 mm/s), the melt pools are visible as tracks (about 100-µm wide), which show semicircular structures (ripples). The ripples are observed due to the height differences and are oriented in the opposite direction with respect to the neighboring tracks (indicated by red arrows). These ripples are defined by their radius and the frequency of their occurrence along the scan track. In addition, there is a high oxide content between the individual tracks and a small number of melt-spatter particles is present on these surfaces. With increasing laser speed at low power, as in Figure 5b (250 W and 1700 mm/s), the tracks do not run in a straight line, but deviate. Gaps form between the individual scanned tracks, in which un-melted powder particles are lodged. The previously scanned, un-melted tracks are exposed in these gaps. No ripples form on the surface of the tracks, but there are cracks and deformations on the surface. At the highest energy density, Figure 5c (370 W and 500 mm/s), the solidified scanned tracks are evenly distributed. Their width increases significantly to more than 300 µm, which leads to the next scanned track overlapping the center of the adjacent melt pool and thus partially obscuring the individual scanned tracks. The frequency and radius of the ripples are increased, and the ripples run almost parallel to the scan direction. With these parameters, the number of spattered melt particles increases. In addition, there are different regions with different textures at the edge of the melt track. Therefore, the analysis of these regions in Figure 6 was carried out in the area where the white square is drawn. At a high scanning speed and high power, as in Figure 5d (370 W and 1700 mm/s), no ripples are seen on the scan tracks, but the number of scattered melt particles is still relatively high. The overlap is small, and the scanned tracks deviate from the straight lines, similar to Figure 5b, but to a lesser extent. There is a large amount of surface deformation and cracks on the surface.

Figure 6a shows a magnified SEM image of the area shown in Figure 5c with the corresponding EDS elemental distribution maps. The SEM image (Figure 6e) of this other texture shows a subsurface dendritic, characteristic tree-like structure of crystals that grow as the molten metal solidifies. In addition, the oxygen content is the lowest in this region, as shown by the EDS elemental distribution map for oxygen in Figure 6b. A higher oxygen content is found at the edge of this dendrite region, where some agglomerated blobs are formed, in the middle of the scanned path and with some local elevations at places where the metal particles are splattered. In contrast to the oxygen distribution on the surface, the distribution of aluminum atoms, shown in Figure 6c, is homogeneous. The distribution of silicon shown in Figure 6d is partially inhomogeneous, as the silicon concentration is lower at the boundaries of the scanned tracks.

3.2. Surface Roughness

Figure 7 shows graphs of the top (Sa, Sdq, Sku) and side (Sa, Sdq, Ssk) surface-roughness parameters as a function of the processing parameters of the laser scanning speed and power. The Ssk on top surfaces and the Sku on side surfaces did not show any obvious trends, so only Sku and Ssk are presented for the different surfaces. The scanning speed is given on the x-axis, and the colors of the graphs represent the laser powers, represented in a legend.

The roughness parameter of the top surfaces (Figure 7a) increases linearly with the speed of the laser scanning and decreases with increasing laser power. At 370 W and 500 mm/s, the highest energy density, and the lowest Sa surface roughness of 2.3 μm is achieved. At the same scanning speed and with a reduction in the laser power to 250 W, the surface roughness increases to 6.5 μm. At the highest laser scanning speed of 1700 mm/s and a power of 370 W, the surface roughness values are similar to 7.2 μm for all except the specimens made with 250 W, where the highest roughness value of 8.2 μm is achieved. The highest gradient of the Sa surface parameters is present at the maximum laser power, whereby the gradient of the surface roughness decreases with the decreasing laser power. The roughness parameter Sa on the side surfaces (Figure 7b) decreases with increasing laser scanning speed or decreasing laser power. At the highest laser power of 370 W and 500 mm/s, Sa is highest at 13.3 µm, while when the laser power is reduced to 250 W at 500 mm/s, the roughness decreases to 10.6 µm. At all the other scanning speeds and powers, Sa is around 9.3 µm, with a slight increase to 10.2 µm at the highest scanning speed. The minimum surface roughness of the side surfaces for all the scanning speeds is achieved at the lowest laser power of 250 W, with the minimum roughness of 8.4 μm achieved at 900 mm/s. When comparing the top surface with the side surfaces, the surface roughness of Sa is 30% lower overall. For the side surfaces, the trend of the effect of the laser scanning speed is reversed compared to the top surfaces. At the lowest scanning speed of 500 mm/s, the roughness Sa is higher than at the higher scanning speeds, where it decreases. The roughness parameter Sdq on top surfaces (Figure 7c) increases linearly with the increase in laser scanning speed and the reduction in laser power. The lowest value of 0.2 is achieved at 370 W and 500 mm/s, and the highest value of 0.68 is achieved at 250 W and 1700 mm/s. The positive gradients in relation to the scanning speed are similar for all the laser powers. The roughness parameter Sdq on the side surfaces (Figure 7d) increases at lower and higher scanning speeds. At a lower scanning speed of 500 mm/s it is 1.15, and at a higher scanning speed of 1700 mm/s it is 0.85. The lowest value of 0.72 is achieved at 370 W and 1300 mm/s. There are no significant differences in the change of laser power, and it shows a parabolic trend. Overall, the top surfaces are smoother than the side surfaces and the parameter Sdq shows similar trends to Sa. The advantage is that the standard deviation is lower, and this parameter is easier to distinguish between laser powers.

On the top surfaces, the Sku roughness parameter in Figure 7e is between 3 and 6 for all the processing parameters. The Sku value increases slightly when the scanning speed is increased, but increases when the laser power is lower. Therefore, the Sku parameter is 3.7 at a high laser power of 370 W and 6.2 at a low laser power of 250 W. This means that the surfaces at lower laser powers are more “spikey”.

On the side surfaces, the Ssk roughness parameter in Figure 7f is positive for all the processing parameters. The value of the Ssk parameter decreases linearly as the scanning speed is increased or the laser power is reduced. At 370 W and 500 mm/s the maximum Ssk value of 0.97 is achieved, while at 250 W and 1700 mm/s the minimum value is 0.4.

The ANOVA results for average surface roughness on both the top and side surfaces, presented in

Table 2. Two-way ANOVA results on average surface roughness, for top and side surfaces, as a function of laser power and scanning speed (SS: Sum of Squares, DoF: Degrees of Freedom, MS: Mean Square, F: Variance ratio, P: Probability and F-crit: Threshold).

Variable	SS	DoF	MS	F	p-Value	F-crit
TOP SURFACE
Laser power	13.05	3	4.35	13.22	0.00119	3.86
Scanning speed	21.93	3	7.31	22.22	0.00017	3.86
Error	2.96	9	0.33
SIDE SURFACE
Laser power	1.88	3	0.62	2.41	0.13375	3.86
Scanning speed	30.03	3	10.01	38.50	1.8 × 10−5	3.86
Error	2.34	9	0.26

, reveal distinct effects of laser power and scanning speed. For the top surface, both laser power and scanning speed show statistically significant interaction between laser power and scanning speed (F > 3.86, p ≈ 0), demonstrating that these processing parameters cannot be considered in isolation. The extremely low p-value and high F-value indicate that the effect of laser power is strongly dependent on the scanning speed level, and conversely, the influence of scanning speed varies substantially with different laser power settings. In contrast, the side surface roughness shows a highly significant effect from scanning speed, with a p-value close to zero, whereas laser power does not exhibit a statistically significant impact, as evidenced by a p-value (p = 0.13375 > 0.05). Overall, scanning speed consistently and strongly affects surface roughness on both top and side surfaces, while laser power plays a significant role only on the top surface.

3.3. Surface Porosity

Figure 8 shows the surface porosity of the printed, polished surfaces for the individual processing parameters of the respective laser power and laser scanning speed. In general, the surface porosity decreases up to a certain point with an increasing energy density (e.g., with an increasing laser power and a decreasing scan speed). The surface porosity decreases from 14.36% to a minimum of 0.14%. As the scanning speed increases, the pore size, geometry and quantity increase, while increasing the laser power reduces the surface porosity to a certain extent.

At a moderate energy density in quadrant I, the pores are relatively small (20 μm), circular and evenly distributed for all the parameters. The surface porosity in this quadrant is around 0.5%. In quadrant IV, which also has a moderate energy density, the porosity is similar, but there are some larger pores (80 μm) randomly distributed on the surface. The specimen at 330 W and 1700 mm/s, which has the lowest energy density in this quadrant, has the highest surface porosity of 2.32%. In quadrant II, the surface porosity begins to increase dramatically with an increasing laser scanning speed, from 1.04% to a maximum value of 14.36% at 250 W and 1700 mm/s. The pores have an irregular shape and can be up to 400 μm in size. Some of the pores are randomly distributed, but some are located on straight lines, as indicated by a red dashed line in Figure 8. In quadrant III. the surface porosity decreases with increasing energy density to 0.14% at 370 W and 900 mm/s. However, when the scanning speed is further reduced and the laser power is increased, the surface porosity increases again. The number of pores increases, as does their size (250 μm). The pores have a circular shape.

Figure 9 shows the surface porosity of the side surfaces for each laser power and scanning speed. In general, the surface porosity is low and decreases with increasing energy density. With respect to the build direction, there are no significant differences in the distribution of the surface porosity. In quadrants I and IV, the surface porosity is the lowest, about 0.12%, and reaches 0.03% of the minimum surface porosity at 250 W and 500 mm/s. In quadrant II, the porosity increases with increasing laser speed and decreasing laser power to 2.63%, where the pores are irregular and randomly distributed with a size of about 150 μm. In quadrant III, the number of pores and their size increases drastically with decreasing scanning speed and laser power. The pores are mostly circular and have a size of up to 300 μm. Compared to the porosity of the top surfaces, the side surfaces in quadrants I, II and IV have a 40% lower surface porosity, but in quadrant III, the porosity on the side surfaces is 40% higher, especially at 500 mm/s.

The ANOVA analysis of porosity (

Table 3. Two-way ANOVA results on surface porosity, for top and side surfaces, as a function of laser power and scanning speed (SS: Sum of Squares, DoF: Degrees of Freedom, MS: Mean Square, F: Variance ratio, P: Probability and F-crit: Threshold).

Variable	SS	DoF	MS	F	p-Value	F-crit
TOP SURFACE
Laser power	42.03	3	14.01	1.53	0.27291	3.86
Scanning speed	82.16	3	27.39	2.99	0.08844	3.86
Error	82.49	9	9.17
SIDE SURFACE
Laser power	2.24	3	0.75	0.44	0.72821	3.86
Scanning speed	6.34	3	2.11	1.25	0.34755	3.86
Error	15.18	9	1.69

) in relation to processing parameters, for both the top and side surfaces, revealed low F-values (F < F-critical) and p-values exceeding 0.05, indicating that neither laser power nor scanning speed had a statistically significant effect on porosity within the tested ranges. However, this does not necessarily imply the absence of a relationship, as ANOVA is limited in detecting nonlinear trends, as can be observed from images, as the method compares mean differences between discrete groups and assumes linear or monotonic effects, potentially missing more complex patterns such as U-shaped or peak responses.

Different pore geometries were observed in certain ranges of the processing parameters. Particularly in quadrant II, especially for the top surfaces, the geometry of the pores increases, and their shape becomes irregular. This type of pore is shown in Figure 10a,b. The pores are about 100 µm-wide and 200 µm-long and have irregular boundaries. These boundaries are relatively smooth, but they have different radii. Partially fused powder particles may be trapped in these pores, some of which have been partially polished. In quadrant III, on the other hand, especially on the side surfaces, the pores on the surface have a circular shape, as can be seen in Figure 10c,d. Here, the geometry of the pores is spherical, but when they are cut, only the cross-section of a sphere is present, which has a diameter of 50 to 300 µm. There are no powder particles inside the pores, but some fluctuations in the solidified melt can be observed inside.

3.4. Microstructure

Figure 11 shows the geometry of the melt pools on the top of the printed, etched surfaces for the individual processing parameters of the respective laser power and laser scan speed. In general, the melt pools have the shape of thin ellipses, some of which are parallel to each other, and others have different orientations. These melt pools represent the width, direction and rotation of the scanned tracks, and overall, their size increases proportionally with the increasing energy density. Nevertheless, differently sized melt pools are present for a specific specimen, with some of the melt pools being elongated in one direction and extended across the entire image, while others are smaller and have an ellipsoidal geometry. In some cases, the melt pools overlap and obscure the geometries of neighboring melt pools. The difference in orientation between two melt pools that are partially covered is 67°, as shown by the yellow lines at 330 W and 500 mm/s.

At a moderate energy density in quadrant I, the width of the melt pools is relatively small, about 150 µm, and the elongated melt pools are divided into smaller ones. The edges of the melt pools are not straight, but curved, which changes to the width of the melt pools along their length. The geometry in quadrant IV is similar to that in I, but with the reduction in energy density in quadrant IV, a small number of pores can be observed, which are located at the edges at the boundary of the melt pools. In quadrant II, the width of the melt pools is the smallest overall, with a size of about 100 µm. Elongated melt pools are subdivided into smaller shapes, with multiple melt-pool orientations present. As the energy density decreases in this quadrant, the number and size of the pores located at the boundaries of multiple melt pools increases. In quadrant III, the melt pools are the widest, about 200 µm-wide, and relatively uniformly elongated. The edges of the melt pools run in a straight, smooth line and are subdivided into smaller pools. At 370 W and 500 mm/s, some pores can be seen within the melt pools.

Figure 12 shows the microstructure, the geometry of the melt pools, the side surfaces for each laser power and scanning speed. In general, all the melt pools have a semi-cylindrical shape with the radius at the bottom aligned with the printing direction, which is marked with a red arrow. The upper part of a cylindrical melt pool is usually covered by 2 or 3 different melt pools. Their size, width and radius of curvature change with the parameters, with the width and radius of the melt pools increasing as the energy density increases.

In quadrants I and IV, the melt ponds are thin and more compacted. Their bottom radius is larger and the boundaries of the melt ponds are thin. In quadrant II, the height and bottom radius of the melt pools increase with increasing laser speed and decreasing laser power. There are also more side melt pool edges. Some pores can be observed forming at the edges of the melt pools. In quadrant III, the boundaries of the melt pools are wide with a further decreasing scan speed and laser power, and the bottom radius decreases. The spherical surface pores are generally located within the melt pools. Compared to the top surfaces, the side surfaces are a cross-section of the melt pools shown above, so the trends in the changes in melt-pool geometry are related. The enlargement of the melt pools on the surfaces leads to an enlargement on the side.

Figure 13 shows SEM images of the edges of the melt pools on the top surface and their representative microstructure for the selected processing parameters (power, speed), which are boundary parameters of the matrix. In general, the microstructure is very fine, with the different phases being only a few nanometers in size. The microstructure of this alloy consists of small cellular dendrites that form in tubes and are typical of this type of process, as reported in [25]. The two different phases of the microstructure are the white phase, which is the eutectic Al-Si phase, and the darker phase, which is the primary α-Al phase, as reported in [10].

The boundaries of the melt pool shown in Figure 13 consist of two laser-scan tracks. There are different zones with heat-induced changes in the morphology of the cellular dendrite microstructure. These zones consist of the microstructure of the inner melt pools, the coarse region and the fragmented region. In the coarse zone and the inner zone, the phases are circular, but in the fragmented zone, the white phase breaks into smaller pieces. In Figure 13a, at a low power of 250 W and a scanning speed of 500 mm/s, the coarse zone is 14-µm wide and the fragmented zone is about 4-µm wide. With the same power and an increase in scanning speed to 1700 mm/s in Figure 13b, both zones shrink to 8 µm and 3 µm, respectively, and the microstructure inside the melt pool becomes even finer. At the highest energy density in Figure 13c, at 370 W and 500 mm/s, the microstructure in the coarse zone becomes coarser, with an increase in width to 22 µm and the fragmented zone to 13 µm. The inner microstructure also increases in size. With the same laser power and an increase in scanning speed to 1700 mm/s in Figure 13d, the coarse zone becomes thinner and finer, about 10 µm-wide, with a smaller fragmented zone of 4 µm and again a finer microstructure inside the melt pool.

Figure 14 shows SEM images of the edges of the melt pools on the side surfaces and their representative microstructures for the selected processing parameters (power, speed) and boundary parameters of the matrix. In general, the microstructure of the columnar dendrites on the side faces is oriented in the same direction as the printing, which is indicated by the red arrow. At each boundary, there are three different zones where the morphology of the microstructure of the columnar dendrite changes. These zones consist of the microstructure inside the melt pools, the coarse region and the fragmented region. In the coarse zone and in the zone inside the melt pool, the phases are columnar, but in the fragmented zone, the white phase breaks apart.

In Figure 14a, at a low power of 250 W and a scanning speed of 500 mm/s, the coarse zone is 10 µm-wide and the fragmented zone is about 4 µm-wide. With the same power and an increase in scanning speed to 1700 mm/s in Figure 14b, both zones are reduced to 8 µm and 3 µm, respectively, and the microstructure inside the melt pool becomes finer. At the highest energy density in Figure 14c, at 370 W and 500 mm/s, the microstructure in the coarse zone has increased in size, from a few nm to µm, with the width of the coarse zone increasing to 18 µm and that of the fragmented zone to 13 µm. The inner microstructure also increased in size. In Figure 14d, at a laser power of 370 W and a scanning speed of 1700 mm/s, the size of the coarse zone is reduced to about 10 µm and the microstructure is finer, with a smaller fragmented zone of 4 µm.

3.5. Hardness

Figure 15 shows graphs of top and side surface hardness as a function of processing parameters of laser scanning speed and power. Scanning speed is given on the x-axis and colors of the graphs represent the laser powers given in the label. On the top surfaces, with an increase in laser scanning speed and a reduction in laser power, the hardness is increasing. At 370 W and 500 mm/s, at the highest energy density, the lowest hardness of 106 HV is achieved. At the same scanning speed with a reduction in laser power to 250 W, the surface hardness increases to 121 HV. At the highest laser-scanning speed of 1700 mm/s, regardless of the laser power, all the surfaces have similar hardness of 124 HV. The maximum hardness of the top surfaces of 127 HV is achieved at 330 W and 1300 mm/s.

Connecting the hardness with the microstructure of the inside of the melt pools, at different processing parameters, given by the SEM images in Figure 13, a trend can be observed. The smaller the cells, the greater the hardness is, which is at higher scanning speeds. On the side surfaces in Figure 15, with an increase in the laser scanning speed and a reduction in laser power, the hardness is increasing. At 370 W and 500 mm/s, at the highest energy density, the lowest hardness of 93 HV is achieved. At the same scanning speed with a reduction in the laser power to 250 W, the surface hardness increases to 112 HV. At the highest laser-scanning speed of 1700 mm/s and for all the laser powers, the hardness is around 121 HV. The maximum hardness of the side surfaces with a value of 123 HV is achieved at the maximum scanning speed of 1700 mm/s and a power of 250 W or 370 W. Hardness values for side surfaces are about 8% lower than the hardness of the top surfaces. Additionally, the maximum hardness on the top surface is 3% higher than that of the side surfaces.

4. Discussion

4.1. Effect of Processing Parameters on Topography

To provide a greater insight into the different aspects of the topography variations based on AM processing, we will discuss several important phenomena: (1) macro surface characteristics of the top and side surfaces, (2) determination of the operational range of the AM processing parameters, (3) effect of base material properties, (4) geometrical tolerances and post-processing, (5) surface formation on the micro scale, (6) surface oxidation layer, and (7) elemental distribution.

(1) In the AM of parts, there are characteristic and different orientations of the surface features and so the topography changes. Usually, a smoother surface is the goal, which can be optimized through AM process parameters, but this differs for the top and side surfaces. In general, to obtain the smoothest possible top-surfaces, upward-facing surfaces should use the print parameters from quadrant III, while side-facing surfaces should be printed with parameters from quadrant II.

During the printing process there is always only a small amount of molten material present, which depends on the energy of the laser, and the absorption of the laser energy in the powder. During the printing process there is always only a small amount of molten material present, dependent on the energy of the laser. However, as the laser moves away, the molten material begins to solidify. This results in the observed wave-like height structure, which is a combination of the melt solidification characteristics and the wetting contact angle (see Figure 16) that shapes the metal in a form to minimize the surface tension [33]. More specifically, the surface formation and the degree of texture are therefore due to two main parameters.

The first one is the melt temperature. If the scanning speed is high (the power has less of an effect), a relatively small amount of energy is available, so that the melting temperatures are low and the viscosity increases [34]. At lower melting temperatures, the wetting contact angle of the molten metal is high [35], so that it forms hemispherical shapes during solidification, also known as balling (Figure 4, most clearly at 250 W and 1700 mm/s, Figure 16a), separated by a hatch distance (in this case 190 μm). This effect is most dominant at the lowest energy density, where the texture peaks of the scan paths become more visible (Figure 2 quadrant II.). However, as the energy density increases, the temperature of the neighboring surfaces and the melt pools increases, allowing the melt to spread evenly due to the lower viscosity and better surface wetting, as reported in [34], which decreases the contact angle, as shown in Figure 4 at 370 W and 500 mm/s and Figure 16b.

The second key effect is the size of the melt pool. At higher energy densities, there is more volumetric melt because there is enough energy and time (high laser power and slow scan speed) to completely melt the powder particles and neighboring surfaces. Therefore, with a constant hatch distance and increased melt-pool size, the overlap between the individual scan paths is greater (Figure 16b). Due to this overlap, each peak of the laser scanner’s path is partially remelted so that only the top part of the peak of the scan path is not remelted, resulting in flatter surfaces (Figure 2 quadrant III. and Figure 4 at 370 W and 500 mm/s), as shown in Figure 16b. At a low laser power, the remelting process can deform the scan path and the topography can become unstable (Figure 5b). Under these conditions, the melt is more likely to merge than to split into several smaller peaks, creating large gaps in between, as shown in Figure 4a. Furthermore, the overlap of the melt between the neighboring scan lines can be increased by decreasing the hatch-distance parameter so that the surface topography becomes flatter. This could be a useful parameter for optimization when using lower laser powers.

In contrast to the top surfaces, the side surfaces, which are perpendicular to the top surface, show less dependence on the printing parameters. However, they have slightly higher peaks in the surface compared to the top surface (Figure 3), as also reported in [14,36,37]. The side surfaces are more homogeneous with randomly distributed agglomerates of metal and adherent powder particles [15,38], as shown in Figure 17. The predominant effect in the formation of the topography of the side surfaces is the adhesion and melting of the surrounding powder particles adjacent to the printed part, as shown in the model in Figure 17.

When the energy density of the laser is low (quadrant II), the number of powder particles adhering to the side surfaces increases. These particles adhere due to partial laser scattering [8] and the thermal conductivity caused by the applied laser energy. At a low energy density, the reflected energy is lower, so that there is not enough energy to melt the powder particles completely, but they are only partially melted on the printed side surfaces. Furthermore, due to the high thermal conductivity [39] of aluminum, a small amount of additional heat is transferred to the powder particles through the side printed surfaces, so that they only adhere to the surface. With an increase in energy density, more energy is scattered onto neighboring powder particles, most of which are completely melted and therefore form larger melt clusters (Figure 2 in quadrant III.). With this large amount of energy, even the conductive heat is high enough to partially melt the powder particles on the printed surface. Therefore, the topography of the side surfaces is indirectly affected by the processing parameters to form top surfaces.

Furthermore, layering effects on side surfaces (Figure 3 line at 250 W and 1300 mm/s), which are a consequence of the build direction, cannot always be observed, as the scan direction between the individual layers is 67° and so the scan paths rarely run parallel with the side surfaces.

(2) The surface topography is also a key factor in determining the range of processing parameters with which the parts can be printed. If the layer thickness is kept constant, the height of the printed surface can exceed the height of the layer thickness for certain processing parameters, due to the balling effect. This can lead to the blade of the re-coater hitting the printed surface and damaging itself or the surface (Figure 4; optical image). On the other hand, with an increasing layer thickness for certain parameters (usually at a low energy density), the gaps between the individual scan tips can become too deep, so that the laser can no longer penetrate through the deposited powder of the next layer (Figure 18), which worsens the mechanical properties [24].

(3) Another aspect to be considered in the effects on the surface topography is the material properties. In particular, this material possesses good fluidity [40], which is also the reason why it is used in die casting [30]. This allows the melt to spread and form a flatter surface topography, especially at higher laser powers. However, the disadvantage of this material is the relatively low laser-absorption coefficient [41], which consequently requires higher laser powers and more energy. In addition, the lower thermal conductivity reduces the side surfaces’ roughness, as the adhering powder particles are reduced. The ideal printing material therefore has a low viscosity when melting, a low contact angle, a relatively low conductivity and a high laser absorption.

(4) The fourth relevant effect of the topography on the geometric tolerances of the printed components that we would like to emphasize is the large differences in the anisotropy of the surface topography on the top and side surfaces (Figure 2 and Figure 3). As observed, the surface topography has different topography textures depending on the build direction, which must be considered during mechanical post-processing. The depth of the post-processing can be changed depending on the printing parameters to completely remove the surface topography. Otherwise, unintended surface porosity could occur, as shown in Figure 18 as dotted green lines.

(5) Formation of ripples is another phenomenon related to AM processing parameters (Figure 5 and Figure 6). At higher energy densities on the scanned track, the formation of ripples begins (Figure 5a,c). The ripple effect occurs when the laser beam melts the powder. If the melting mode is a keyhole mode [42], a cavity can be created by plasma formation, as shown in Figure 19a. This cavity is surrounded by molten metal, which applies forces to the cavity due to surface tension and hydrostatic pressure [43]. As the laser beam travels along its assigned path, material flows into the cavity, causing metal movement and material spatter (Figure 19b). The metal then solidifies in the opposite direction to the laser movement in the characteristic ripple pattern that can also be seen during welding [44] or numerical simulations [45]. The shape and frequency of the ripples depend on the formation of the plasma and the geometry of the molten pool, which is influenced by the energy density, with larger radii and higher frequency at high energy densities, as also reported in [46]. In addition, the ripple formation of the neighboring scan tracks is in opposite direction, which corresponds to the laser scan pattern of an alternating scan direction, as shown by two red lines in Figure 5a. Furthermore, with this processing parameters and material, cracks are formed on the surface due to surface contraction and deformation due to high thermal gradients (Figure 5b,d). The deformation of the oxide layer is mainly caused by the transfer of turbulence stresses from the melt to the topmost oxide layer formed [47] (Figure 20a). In addition, the shrinkage of the melt due to the different thermal expansion coefficients compared to the oxide layer leads to stresses that cause cracks between the oxide layer and the molten metal [48] (Figure 20a). Therefore, with a detailed look at the surface formation, it is possible to determine which machining processing parameters were used to produce the part, depending on the surface features formed.

(6) Although the printing chamber is filled with argon gas, there are surface oxide layers of varying thickness (Figure 5). A particularly thick oxide layer forms on spattered molten particles that pass through the printing chamber (Figure 6b). This small amount of oxygen comes from the powder itself, which has already been oxidized [39], as the aluminum alloy has a very high affinity for oxidation [49], and is therefore released during melting. A thin oxide layer forms at higher energy densities (Figure 6), where there are high temperatures, and for a very short time, therefore, there is limited oxygen supply, as reported in [48]. However, when cooling is faster and at lower temperatures, the thickness of the oxide layer increases (Figure 5). In addition, extremely thin oxide layers are found near the rescanned laser paths at high energy densities (Figure 6) and are caused by another neighboring rescan that causes an increase in the temperature due to thermal conductivity. In these regions, the oxide layer is melted and molten oxides aggregate, as shown by the EDS (Figure 6a,e). This ultra-thin oxide layer exposes the underlying dendrites (Figure 6e), which resemble the casting process [30]. Compared to the polished, etched surfaces showing the inner microstructure (Figure 13 and Figure 14), which exhibits typical fine cellular dendrites, we could assume that the upper surface cools much more slowly.

(7) In metallic materials, we typically expect a homogeneous distribution of alloying elements, as demonstrated by the uniform dispersion of aluminum atoms across the surface (Figure 6c). This homogeneity arises naturally since aluminum constitutes over 88% of the alloy’s weight. However, silicon exhibits markedly different behavior, with clear aggregation patterns visible in Figure 6d. This inhomogeneous Si distribution likely stems from multiple factors: its higher atomic mass promotes segregation during solidification, driving accumulation in regions of lower surface topography (Figure 20b)—specifically, the characteristic valleys formed at melt track edges. Additional mechanisms such as solidification partitioning and Marangoni flow may further influence this segregation behavior. These observations suggest that in alloys featuring significant atomic mass disparities and pronounced surface topography, elemental distribution may deviate substantially from ideal homogeneity.

4.2. Topography Characterization Using Multiple Parameters

The most common characterization of roughness properties is the Sa value, which provides global deviations of the surface heights from the mean. In doing so, it is clear that the lower surface roughness derives from the top surfaces at higher energy densities, with a minimum Sa value of 2.3 μm obtained in this study, which is consistent with other reports [2,36] (Figure 7). At high scanning speeds, energy can be saved by using lower laser powers, as the power has little effect on roughness at these speeds (Figure 6a). On the side surfaces (Figure 6), the minimum Sa value achieved is 8.2 μm, with reversed parameter trends towards the top surface, as also reported in [50]. The side surfaces are almost three times rougher.

However, due to the specific textural and topographic features of the AM specimens mentioned above, additional roughness parameters such as Sdq, Ssk and Sku are very valuable to better characterize the surfaces. As can be seen in Figure 4 and Figure 16, the top surface is related to the wetting angle, which defines the slopes and curvatures of the surfaces. It is therefore directly related to the parameters Sdq (which describes the smoothness of the surface angle) and Sku (which describes the shape of the height probability distribution). This is also evident in the results where a high energy density smooths the surfaces, allowing Sdq and Sku (Figure 7c,e) to link the formation of surface topography to characterization. According to our results, Sdq is the most robust parameter for the characterization of printed surfaces.

On the other hand, side surface topography is influenced by the ratio of the adhering and fused powder particles, as already described in (Figure 3 and Figure 17). This change in mechanism also requires a change in the surface-roughness parameter, where both Sku and, to a lesser extent, Sdq did not show clear trends (Figure 7d). Therefore, for the side surfaces, the Ssk parameter resulted in the clearest correlation between the processing parameters and side-surface topography (Figure 7f). As the energy density increases, the powder particles form larger melt clusters and increase Ssk, but at lower densities, only partially melted particles are present, making the surfaces flatter but with more valleys, which decreases Ssk.

4.3. Effect of Processing Parameters on Porosity

In contrast to surface roughness, surface porosity (Figure 8 and Figure 9) shows non-linear trends in the change with the processing parameters. The nonlinear effect of printing parameters on porosity is caused by the altered mechanism of pore formation, which changes from keyhole pores at higher energy densities to a lack of fusion and the balling effect in the other spectrum at low energy densities, as also described in [40].

When the pores form due to the keyhole effect (Figure 19) (Figure 8 and Figure 9 in quadrant III) and when metal evaporates during the formation of plasma at higher temperatures, the formed gas becomes trapped as bubbles within the melt, due to melt movement [51]. For this reason, they have a round shape in cross-section (Figure 10b). The increase in this type of surface porosity on the side surfaces (Figure 9) at higher energy densities is due to the cross-section of several printing layers, the larger shape of the melt pools and the pores being located inside the melt pools (Figure 12).

In contrast, the other mechanism of pore formation is for the irregularly shaped pores (Figure 10a,b), (Figure 8 and Figure 9 in quadrant II), which form due to the lack of fusion between the scanned tracks [52]. They are not located within the melt pool, but at its boundaries (Figure 11, quadrant II), and their size increases with decreasing energy density, resulting in larger gaps between the scanned tracks. These pores are also larger than the gas pores and can contain unmelted powder particles (Figure 10a). The pore boundaries are still relatively smooth (Figure 10a), but as observed, they are surrounded by different parts of the melt pools, which is why their boundary radius changes. On the side surfaces (Figure 12), the amount of this type of pores is smaller, since the vertical overlap between the melt pools is greater than in the horizontal layer.

Overall, the minimum surface porosity on the diagonal of the median of the energy densities (quadrants I and IV) is achievable for this set of parameters, with the side surfaces having a lower surface porosity, but the evaporation temperatures of the material should be considered when printing at higher energies.

4.4. Effect of Processing Parameters on Microstructure

Optical microstructure of the etched specimens provided the means for changes in the geometry of the melt pools and their overlap. Ideally, only one scanning direction and one size of melt pools should be present on the top surfaces for each processing parameter. However, on the etched surfaces shown, there are different orientations of the scanning tracks and sizes of the melt pools (Figure 11). The change in the size of the melt pools or the variation of the thickness corresponds to the change in the waviness of the melt-pool depth, which is influenced by the topography of the previously printed layer (Figure 2) and by changes in the processing parameters, as reported in [11,53] for the material 316 L. In addition, the width of the melt pools can be influenced by grinding and polishing, as the melt pools can be sliced at the lower part, which is narrower (Figure 12 and Figure 21a).

The different orientations of the melt pools are due to observations of underneath melt pools (viewing multiple layers) due to polishing the surface at an angle, which is not perfectly parallel to the printing layer; therefore, different printed layers are present. Rotation of melt pools is also present (Figure 11), because when printing, there is 67° of scanning rotation change between each layer. Additionally, melt pools that are sliced at the bottom, which is narrower are therefore exposing underneath melt pools. It should be noted that with grinding angle, microstructure at different heights of the melt pools can be observed.

The cross-sections of the melt pools, the direction of the laser scanning and their overlap with neighboring paths are shown on etched side surfaces (Figure 12). The direction of the scan is indicated when one side of the pool boundary disappears because it is covered by the next adjacent scan. This means that the laser moves in this direction, opposite to the covered boundary. Layer-by-layer overlap in the Z-axis is observed when only the lower part of each melt pool is present. The upper part of the melt pools cannot be observed as it has already been melted again and covered by the next printed layer. With the same parameters, the melt pools should look identical on the entire surface, but due to the rotation of the scanner by 67° in each layer, the cross-section of the melt pools is not always perpendicular to the observed side surface and therefore changes shape and size according to the projection angle (Figure 21a).

Microstructural information can also help to adjust the hatch spacing and layer thickness in relation to the processing parameters to achieve better final properties and ensure a suitable bond between the melt pools. In addition, the texture on the side surfaces is more oriented in the build-up direction, which is why the mechanical properties are anisotropic in powder-bed laser printing.

Consequently, with a higher input-energy density and the same layer thickness, there are fewer boundaries with heat-affected zones in which the microstructure changes (Figure 13), because the size of the melt pools increases (Figure 11, quadrant III). At a high laser power, however, the width of the heat-affected zone increases (Figure 13a,c). The typical cellular dendrite microstructure is present here (Figure 21b), consisting of primary α-Al and Al-Si eutectic [10], with the geometry of the cells depending on the cooling rate and the direction of growth [8]. Therefore, at high energy densities (quadrant III), the cellular microstructure becomes coarser due to slower cooling, but as the laser-scanning speed increases, the cooling becomes faster and the cells become smaller. Moreover, besides the coarse zone (Figure 13), there is another region with a fragmented Si-eutectic phase, which looks like a typical example of a partially aged or solution annealed microstructure, as reported in [26]. Consequently, the re-melting of the previously scanned printed paths leads to some regions already exhibiting microstructural features after aging.

The difference between the side surfaces (Figure 14) and top surface is the change in the geometry of the cells, which change from cellular to columnar, as also reported in [23]. This is due to the orientation of the dendrites, which lie in the same direction as the highest temperature gradient, i.e., in the direction of the previous layer, which is remelted, towards the upper part of the melt pool (Figure 21b). The dendrites solidify in a tubular shape and have a circular shape at the top and a rectangular shape at the side due to the cross-section, which leads to anisotropic properties of the printed parts.

4.5. Effect of Processing Parameters on Hardness

The results of the hardness measurements correspond to the microstructural changes that are influenced by the processing parameters. The overall hardness of the surfaces is around 120 HV (Figure 15), which is typical for this type of process and material [54,55]. The lowest hardness is achieved at the highest energy density, which is related to the largest geometry of the cell sizes especially in quadrant III (Figure 13 and Figure 14). This is caused by relatively slow cooling and an increase in the fragmented zone, which further reduces the hardness, as also reported in [7,24,55] for thermal aging. As the scanning speed increases, the cell size decreases, and thus the surface hardness increases, but in quadrant II, the hardness starts to decrease due to the increased surface porosity (Figure 8 and Figure 9). Overall, hardness exhibits a dual dependence on porosity and cellular microstructure, with competing effects in different parameter regimes. In quadrant II, high porosity dominates, reducing hardness despite fine cell structures. Conversely, in quadrant III, coarse cells and fragmented zones lower hardness even with minimal porosity. Combining both effects is evident in quadrant IV, where top surfaces achieve the highest hardness through optimal balance of fine microstructure and low porosity. The 8% hardness reduction on side surfaces further confirms microstructural effects, though precisely decoupling porosity and cell size contributions remains challenging due to their coupled formation during solidification.

5. Conclusions

• With a high laser power and a slow scanning speed, top surfaces become flatter with more overlap due to sufficient spread of melt. With an increase in the scanning speed, scanned paths become more distinguished with increased height and with less power, the balling effect becomes predominant, causing waviness and larger gaps in between scanning lines. Side surfaces are mostly made up of adhered powder particles, caused by thermal conductivity or laser dispersion and are more predominant at lower energy densities.
• The top surface roughness is, in general, three times lower than on the side surfaces, where a minimum Sa of 2.3 μm is achieved at 370 W and 500 mm/s for the top and 8.2 μm at 250 W 1300 mm/s on the side. With an increase in energy density, the roughness for top surfaces decreases, while for the side, it increases. Supplementary parameters of Sdq, Sku and Ssk can be used, to additionally describe the surface roughness and anisotropy.
• The surface porosity of the printed specimens shows circular gas pores formed in the region at higher energy densities. On the other hand, a lack of fusion pores, with irregular shapes, filled with powder particles, are present at lower energy density. The porosity of the top surfaces is higher than that of the side surfaces, except in the gas-formed region of pores, where the porosity is higher on the side surfaces.
• The melt-pool geometry with an increase in the energy density is increasing, showing the orientation and scanning direction on the top surfaces and their cross-sections on the side surfaces. Additionally, the heat-affected zones at the boundaries comprise three regions, where the area is increasing with energy density. With a decrease in the energy density the cellular dendritic microstructure is becoming smaller. The side surfaces show the same type of heat-affected zones and size changes in the dendrites, but the dendrites have columnar geometries instead of cellular.
• The hardness mostly depends on the geometry of cell dendrites, which are affected by cooling rates. At low scanning speeds and high laser powers, the cooling rate is slower and the cell size increases, therefore reducing the hardness. With an increase in the scanning speed, the cells become finer and the hardness is increased, but this is limited by the higher porosity, especially at lower laser powers. On the other hand, side surfaces have a lower hardness due to the larger columnar size of the dendrites, but have less surface porosity, and therefore the hardness is constantly increasing even at lower energy densities.
• By using ANOVA analyses it was found that scanning speed consistently and strongly affects surface roughness on both top and side surfaces, while laser power plays a significant role only on the top surface. Porosity, however, does not show a clear linear relation to laser speed or power, but it has probably more complex behaviors.